Pathological stiffening by crosslinking glycation of titin

35 Heterogeneous, non-enzymatic glycation chemistry triggered by s ugar-derived metabolites is typical of diseases 36 that also entail pathological stiffening of cells, such as diabetes and age-related disorders. However, the mechanisms 37 responsible for cell stiffening and the role of glycated biomolecules remain largely unexplored. Here, we show that 38 glycation of cardiac titin, a giant intracellular protein scaff olding contractile sarcomeres, is increased in diabetes 39 and leads to rigidification of b oth the protein and cardiomyocy tes. Mechanistically, glycation-induced titin 40 stiffening results from decreased contour length and enhanced f olding of otherwise structurally intact protein 41 domains following extensive formation of intramolecular crossli nking advanced glycation end products (AGEs). 42 These stiffening effects outweigh softening contributions by competing, non-crosslinking AGEs. In combination, 43 our work overcomes the intrinsic chemical complexity typical of glycation to uncover crosslinking AGEs as a 44 source of pathological stiffening of cells, which we propose co ntributes to tissue dysfunction in situations of 45 glycative stress. 46 47 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 3

48 Mechanics play inextricable roles in biological processes like development or stem cell differentiation1, and, when 49 dysregulated, contribute to the onset and progression of disease. For instance, stiffening of the extracellular matrix 50 (ECM) is linked to the growth of cancer cells 2, and reduced compliance of the cardiac walls leads to heart fa ilure 51 because ventricles cannot properly accommodate blood during dia stole3. Based on these observations, drugs 52 countering mechanical alterations are gaining increasing traction in the clinic4,5. However, progress is hampered by 53 an incomplete understanding of the molecular mechanisms underly ing pathogenic changes in tissue mechanics. 54 Here, we address why diabetic tissues and cells are stiffer tha n healthy counterparts, a situation that contributes to 55 increased risk of heart failure among other systemic complications6-8 (Figure 1A). 56 The macroscopic mechanical properties of tissues derive from the nanomechanics of their constituent biomolecules. 57 This is best exemplified in striated muscle, where both extra- and intracellular proteins are well-known to set tissue 58 mechanics9. Key players include collagens 10, the main structural scaffold of the ECM, and the giant intrac ellular 59 protein titin11-13, a major contributor to cardiomyocyte (and myocardial) stiffne ss at physiological strains9 (Figure 60 1B). Considering this emergence across scales, it comes as no sur prise that modulation of protein nanomechanics, 61 for instance through protein isoform switch or by posttranslational modifications (PTMs), represents a physiological 62 regulator of global tissue mechanics 11,14. Importantly, in some diseases these very same mechanisms can result in 63 pathological maladaptation perturbing tissue homeostasis. Such detrimental effects can be especially relevant in the 64 case of PTMs favored during events of chemical stress, as exemp lified by oxidations triggered by 65 ischemia/reperfusion damage15,16. The mechanical consequences of other forms of chemical stress, including those 66 induced by glycation typical of diabetes and age-related diseases 17,18, remain incompletely understood. This so-67 called glycative stress results from abnormally high concentrat ions of sugars and/or related metabolites, which 68 induce a complex cascade of non-enzymatic reactivity between th eir carbonyl groups and amino or guanidino 69 groups of both extra- and intracellular proteins. These reactions eventually form a heterogeneous set of compounds 70 known as advanced glycation end products (AGEs) 19, which includes both crosslinking and non-crosslinking 71 modifications. Most often, many different AGEs coexist in affec ted tissues, challenging integrative functional 72 characterization19. 73 The main intracellular glycating compound is methylglyoxal (MG), a small and highly electrophilic -oxoaldehyde 74 formed as a side product of glycolysis 20,21. MG is thousands of times more reactive than glucose 22; thus, it can 75 rapidly and extensively modify proteins even in the low M concentration found in cells 22,23. MG reacts with 76 arginine side chains to form, among others, MG-derived hydroimidazolones (MG-H1, MG-H2 and MG-H3)23, and 77 with lysine side chains, mainly yielding Nε-carboxyethyllysine (CEL)24 and a crosslinking MG-derived lysine dimer 78 (MOLD)25. 79 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 4 Despite the concurrence of glycative stress and tissue stiffening, the causal relationship between them has remained 80 elusive (Figure 1A). While crosslinking AGEs targeting ECM proteins have been pro posed to contribute to tissue 81 rigidification6,26-28, it remains unclear what the effect of competing, non-crosslinking counterparts may be since this 82 type of modifications is expected to induce protein softening 29. More importantly, mechanical alterations of the 83 ECM cannot explain why diabetic or aged cells themselves are st iffer, as observed with endothelial cells 30,31 and 84 cardiomyocytes6,32,33. Building on circumstantial observations that titin is glycate d in diabetic and aged 85 myocardium34,35, we hypothesized that titin glycation could contribute to card iomyocyte stiffening in these 86 conditions. To test this hypothesis, here we have exploited single-molecule force spectroscopy, mass spectrometry, 87 computer simulations, high-resolution NMR and cell mechanics ex periments to demonstrate that AGEs in titin 88 stiffen cardiomyocytes in situations of glycative stress. 89

90 Increased glycation products in titin from diabetic myocardium 91 To study whether titin glycation plays a role in pathological c ardiomyocyte stiffening, we first set out to provide 92 additional support to available evidence indicating that cardia c titin glycation increases in conditions inducing 93 glycative stress34,35. To this aim, we applied low-percentage sodium dodecyl sulfate-polyacrylamide electrophoresis 94 (SDS-PAGE) to protein extracts obtained from the myocardium of 20-week-old ob/ob mice, a model of type II 95 diabetes induced by obesity 36, and from WT controls. Next, we sliced bands corresponding to titin and digested 96 samples with chymotrypsin. Ensuing mass spectrometry (MS) analy sis identifies three times more peptides 97 containing CEL, MG-H and N 6-carboxymethyllysine (CML) AGEs in ob/ob animals than controls ( Figures 1C, 98 S1A). Similarly, we find 3.5-fold enrichment in glycated titin peptides in human myocardium from type II diabetic 99 patients compared to non-diabetic counterparts (Figure 1D, S1B). 100 101 Cardiomyocyte stiffening by methylglyoxal 102 Our MS data confirm that pathological glycative stress during diabetes results in titin glycation, a modification that 103 could contribute to cardiomyocyte stiffening 6,33. To examine this possibility, we determined the passive stiffnes s 104 of skinned cardiomyocytes undergoing glycative stress induced by MG. In a first approach, we used nanoindentation 105 by atomic force microscopy (AFM), a method that measures transv erse stiffness of specimens from force-distance 106 (FD) curves that relate cantilever force to vertical displaceme nt during indentation (Figure 1E,F ). We find that 107 typical FD curves recorded on neonatal murine cardiomyocytes treated with 50 mM MG for 4 h at 37ºC have a 108 steeper slope at the contact region compared with controls, ind icative of MG-induced cell stiffening ( Figure 1F). 109 Quantification confirms a two-fol d increase in the Young’s modu lus of MG-treated cells (19±4 kPa, mean±SEM 110 unless stated otherwise) compared to controls (7±2 kPa) (Figure 1G). 111 112 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 5 Next, we followed a complementary approach to assess the longit udinal stiffness of MG-treated cardiomyocytes. 113 To this end, we tethered single skinned rat cardiomyocytes to a motor and a force transducer and applied a tensile 114 test consisting of six stepwise length increases of 5% the init ial length (L0). Each straining step causes a sharp rise 115 in passive force, followed by a slow relaxation phase as expected from the viscoelastic nature of cardiomyocytes29. 116 After this initial test, we incubated cells at slack length with buffer including or not 50 mM MG for 30 min at room 117 temperature (RT), and the tensile test was repeated. To average data, we normalized passive force signals for each 118 cell according to values at 1.3 L0 in the first tensile test (Figure 1H). Results indicate that MG treatment induces a 119 90% increase in passive force at 1.3 L 0 from 5.0±0.8 kPa to 9.5±2.6 kPa, while incubations in the absen ce of MG 120 show no effect (Figure 1H-J). Importantly, additional control experiments demonstrate that these stiffening effects 121 do not result from potential MG-induced modification of the fre e amine group in ATP, a molecule present in the 122 experimental buffers to prevent the formation of actin-myosin crossbridges37. In these control experiments, we find 123 that relaxing buffer containing ATP preincubated with MG does not noticeably affect the passive stiffness of 124 cardiomyocytes, in agreement with preserved blockage of actin-myosin interactions (Figure S2). 125 126 In combination, our cell mechanic s experiments demonstrate that MG-induced glycation leads to cardiomyocyte 127 stiffening in both transverse and longitudinal directions, in a greement with a recent report using skeletal muscle 128 fibers38. Considering that titin is the main contributor to passive sti ffness of cardiomyocytes, particularly in the 129 longitudinal direction 39, our data support that the increased glycation of cardiac titi n in diabetes is indeed a 130 contributor to cardiomyocyte rigidification. To mechanistically scrutinize this possibility, we examined the effects 131 of MG-induced glycation on titin mechanics. 132 Extensive formation of intradomain crosslinking AGEs in titin I91 domain following incubation with MG 133 The contribution of titin to cardiomyocyte stiffness results from the mechanical extension and contraction dynamics 134 of the I-band domains of the protein40 (Figure 1B). Among them, the N2Bus and PEVK regions behave mostly as 135 random coil/entropic springs t hat easily adapt their length to the pulling force in an elastic manner, while folded 136 immunoglobulin-like (Ig) domains are more rigid because they sequester residues that do not contribute to the 137 protein’s contour length. However, Ig domains can unfold upon application of force, which softens titin as a result 138 of the associated increase in contour length ( Figure 1B). This mechanical unfolding is reversible at forces lower 139 than 10 pN41. Modulation of mechanical unfo lding/folding transition kinetics in titin domains can result in large-140 scale changes in the mechanics of the full-length protein, typi cally with opposing outcomes for crosslinking and 141 non-crosslinking modifications, as observed with oxidative modifications29,42,43. 142 To determine the mechanical effects of MG-induced glycation on titin Ig domains, we initially focused on the I91 143 Ig domain of the protein (also known as I27, PDB code 1TIT). I91 contains eight solvent-exposed lysine residues 144 that can be modified by MG, potentially resulting in both cross linking and non-crosslinking AGEs (Figure 2A ). 145 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 6 We first incubated purified (I91)8, a tandem repeat polyprotein containing eight domains of I91, in the presence of 146 50 mM MG for 72 h at 37°C and evaluated results using size-exclusion chromatography (SEC) ( Figures 2B,C, 147 S3A). We used two control samples, i.e. untreated, pristine (I91) 8, and (I91)8 incubated in the absence of MG. In 148 the three conditions, we observe a peak at 15 mL elution volume marking monomeric (I91) 8 (Figures 2C, S3B). 149 Protein aggregation was evident in the sample incubated without MG, resulting in lower signal intensity for 150 monomeric (I91)8 in this preparation. In contrast, protein aggregation was not detected in MG-treated (I91)8; in this 151 case, the main (I91)8 peak is broader with higher contribution of species at lower elution volumes as expected from 152 some degree of MG-induced intermolecular crosslinking. Unreacte d MG was effectively separated from (I91) 8, as 153 indicated by the prominent peak appearing at the exclusion volu me (20-22 mL). To examine the extent of 154 modification by MG, we evaluated (I91)8 samples using MALDI-TOF/TOF mass spectrometry. While the spectrum 155 of pristine (I91) 8 shows a well-defined, narrow peak centered at 81.5 kDa matchin g the theoretical mass of the 156 protein (81.7 kDa), the sample treated with MG yields a much broader peak at 87.6 kDa indicating the presence of 157 a mixture of (I91) 8 molecules containing different extents of MG-derived AGEs (Figure 2D ). Accordingly, the 158 band corresponding to MG-treated (I91)8 in SDS-PAGE gel is quite diffuse (Figure S3B). 159 Having proved that (I91) 8 is extensively modified by MG, we used single-molecule force-spectroscopy by AFM 160 (here referred to as AFS) 44 to characterize the mechanical effects of glycation on I91. We subjected single (I91) 8 161 proteins to a 40 pN s-1 increment in pulling force between 0-260 pN, while simultaneously measuring protein length 162 (Figure 2E-G). As expected, mechanical unfolding of pristine (I91) 8 or (I91)8 incubated in the absence of MG 163

in stepwise ~25 nm increases in length at forces around 150-170 pN, where every step corresponds to a 164 single I91 unfolding event ( Figure 2E,F )29. Since our single-molecule experiments rely on non-specific 165 physisorption of proteins to the AFM cantilever, we find a variable number of 25-nm events in individual recordings 166 (seven and six in the examples in Figure 2E,F, respectively). Notably, many steps with shorter lengths are evident 167 in single-molecule recordings obtained with the MG-treated (I91 )8 sample (Figure 2G, short steps marked by red 168 stars), suggesting that many unfolding domains contain covalent bonds that prevent full mechanical extension of 169 the polypeptide, as previously observed for disulfide and isopeptide bonds45,46. 170 To quantify the extent of MG-der ived crosslinking modifications in an unbiased manner, we selected single-171 molecule recordings following increasingly stringent fingerprin ting criteria44. Initially, using a lax criterion, we 172 extracted the step size and unfolding force of all events in traces containing ≥2 events with at least one being 25 nm 173 in length, and represented the results in bidimensional histogr ams. For pristine (I91) 8, this analysis identifies the 174 expected main population of events centered at 25 nm that emerges from a background of non-specific events typical 175 of single-protein AFS measurements44 (Figure 2H, 25 nm population shaded in blue). Similar results are obtaine d 176 with control (I91) 8 incubated in the absence of MG ( Figure 2I ). In contrast, the abundance of ~25 nm events 177 considerably drops in the sample treated with MG, where a new population of events between 9-14 nm is now 178 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 7 observed (Figure 2J; population of short steps shaded in light red). For precise estimation of the proportion of these 179 shorter steps, we followed a stricter fingerprint criterion ana lyzing only traces that contain 9-14 nm and/or 23-27 180 nm steps (Figure 2K-M). We find that 77% of events in the MG-treated (I91) 8 sample correspond to short steps, 181 strongly suggesting that most I91 domains in these conditions contain MG-induced covalent crosslinking 182 modifications. By setting incubations with MG of different dura tion, we find that the proportion of short steps 183 reaches saturation after 48 h of reaction ( Figures 2N, S4). Importantly, these experiments also reveal that the sole 184 addition of MG induces intradomain crosslinks in 24% of I91 dom ains despite its immediate removal by SEC, a 185 process that is completed in less than 1 h at 4ºC (open circle in Figure 2N). This result indicates that MG quickly 186 induces a subset of covalent crosslinking modifications in the I91 domain of titin. 187 Overall, our single-molecule experiments with MG-treated (I91) 8 indicate that intradomain covalent crosslinking 188 modifications are very prevalent and can target the majority of I91 domains despite expected chemical competition 189 with non-crosslinking modifications. 190 Detection of crosslinking modification MOLD in MG-treated I91 191 We sought to characterize the chemical nature of the crosslinking AGEs formed in MG-treated I91. With that aim, 192 we first produced glycated and control (I91) 1 preparations ( Figures 3A, S5, Text S1 ). Similar to (I91) 8, the 193 molecular mass of (I91)1 determined by MALDI-TOF/TOF increases upon 24 h incubation wit h 50 mM MG, and 194 the peak corresponding to glycated (I91) 1 is broader than that of the pristine protein ( Figure S6). This indicates 195 equivalent extent and heterogeneity of glycation in (I91)1 and (I91)8. Next, we collected the 15N,1H-TOCSY-HSQC 196 NMR spectra of 13C,15N-(I91)1 samples. While the projection of the H-H planes in the spectrum for non-glycated 197 (I91)1 displays well-defined cross-peaks at 7.2 ppm correlating lysine-H with the other intra-residue protons, these 198 signals disappear in MG-treated (I91)1, which proves that MG modifies the NH  group of the lysine side chains of 199 (I91)1 (Figure 3A,B). 200 201 To investigate the chemical nature of the AGEs formed in glycat ed I91, we incubated unlabeled (I91) 1 with 13C-202 labeled MG. The advantage of this strategy is that only carbons coming from MG become NMR-visible, thus 203 allowing straightforward comparison with commercially available chemical standards of AGEs. The 1H,13C-HMBC 204 spectra of unlabeled (I91)1 modified with 13C-MG evidence the correlation of some aromatic (i.e. between 6 and 9 205 ppm in the 1H dimension) one bond C-H cross-peaks with other C-H cross-peak s. This demonstrates that MG 206 induces the formation of aromatic AGEs on I91 (Figure 3C) . To assess whether these si gnals could arise from 207 MOLD, we collected the 1H,13C-HMBC spectrum of MOLD standard (sMOLD) and carried out its ch emical shift 208 assignment (Table S1). The 1H,13C-HMBC signals obtained for sMOLD match those observed for glyc ated (I91)1 209 (Figure 3C), unequivocally demonstrating the formation of MOLD on MG-treated I91. Subtle shifts in the 1H,13C-210 cross-peaks of MOLD formed on I91 compared to sMOLD are likely due to the different chemical environments. 211 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 8 In addition, the low intensity peaks appearing next to the C5-H 2, C5-H4 and C2-H4 cross-peaks suggest that 13C-212 MG-treated (I91) 1 might contain at least two different MOLD moieties with differ ent protein chemical 213 environments. 214 215 In combination, our NMR data confirm that modification of I91 l ysines by MG is substantial and identify MOLD 216 as a crosslinking AGE present in the modified protein. 217 218 Restricted conformational flexibility and preserved fold of glycated I91 219 Our interpretation of the AFS data assumes that glycated I91 remains folded in the absence of force application. To 220 obtain independent validation of this assumption, we further used NMR spectroscopy to get insights at the residue 221 level on potential structural effects in glycated I91. We obser ve that the positions of the structural fingerprint 15N-222 HSQC amide cross-peaks in (I91) 1 do not change upon incubation with MG ( Figure 3D), a first indication that 223 glycation does not induce major alterations to the structure of I91. Furthermore, we do not detect any relevant 224 change in the ring shifted resonance 1H signal of L58- CH3, which arises from its structural proximity to the 225 aromatic side chain of W34 in the hydrophobic core of the domai n (Figure S7A,B), nor in the long-range NOE 226 signals that result from ter tiary structural contacts (Figure S7C ). Using the backbone chemical shifts (i.e. N, H N, 227 Cα, C, Hα and CO) assigned for all residues in 13C,15N-(I91)1 incubated in the presence or absence of MG (BMRB 228 codes 53390 and 53389, respectively), we also observe that glycation does not affect the β-sheet propensity scores 229 of the β-sheets of the domain ( Figure S7D-F). Altogether, these data prove that glycation mediated by MG d oes 230 not impact the secondary nor the tertiary structure of I91. 231 Despite preserving the global fold of the I91 domain, MG-treated samples show increased intensity of many HSQC 232 peaks (Figure 3E). This suggests that glycation shifts I91 towards a predominan t conformation, thereby reducing 233 exchange between minor conformational populations. The most aff ected residues are those mainly located at the 234 ABED β-sheet, as well as in the loops connecting these strands and in those connecting them with the A’GFC β-235 sheet (Figure S8A). Intriguingly, the HSQC peaks that mostly change their intens ities upon glycation are located 236 at the N-terminus of strand D and at the C-terminus of strand E , which could indicate modification of nearby C47 237 and C63 (Figure S8A). However, the chemical shifts of the C β of C47 and C63 (highly sensitive to the oxidation 238 state of the thiol group) did not remarkably change upon glycation (27.3 ppm for C47 and 28.1 ppm for C63, Figure 239 S8B). This observation aligns with previous results, which show that both cysteine residues in I91 have low solvent 240 accessibility when the domain is folded, and therefore cannot b e targeted by reagents in the solution at 37°C 29 241 (Figure S8C). 242 To evaluate the effect of glycation on the dynamics of I91, we acquired NMR relaxation data (i.e. R1, R2 and 15N 243 HET-NOE; Figure S9) and determined the amplitudes of the conformational fluctuations of the backbone amide 244 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 9 groups in terms of order parameters (O2, on the ps to ns time scale) according to a model-free formalism47. O2 values 245 were calculated considering that the best-fit rotational diffusion tensor is anisotropic with a correlation time (c) of 246 10.6 and 11.5 ns for untreated and MG-treated (I91) 1, respectively (Figure 3F ). The mean values (±SEM) of the 247 backbone order parameters calculated for the L1-L89 stretch ( Text S1) are 0.701±0.011 for I91incubated in 248 the absence of MG and 0.768±0.013 for MG-treated I91 (p<0.0001 for paired samples t-test), suggesting that 249 glycation restricts overall protein conformational dynamics and rigidifies the domain. This result is even clearer 250 from the comparison of O2 (i.e. ΔO2= O2I91 glycated - O2I91) at the residue level ( Figure 3G,H). Remarkably, not all 251 regions showing increased O2 are located near MG-sensitive lysine residues, which indicates that glycation can 252 reduce structural fluctuations in an allosteric manner. Finally, the restricted conformational flexibility of glycated 253 I91 was also captured by a reduced conformational entropy of th e protein backbone (i.e. ΔSconf=-114.2±2.9 Jꞏmol-254 1ꞏK-1 or TΔSconf=-34±0.8 kJꞏmol-1 at 298 K; error is SD), as calculated from NMR relaxation data48. 255 256 In short, structural characterization of glycated I91 proves that MG does not induce any major change to the fold of 257 the protein. However, we detect reduced conformational flexibility of several regions of the glycated domain, which 258 probably results from the formation of crosslinking AGEs captured in single-molecule force-spectroscopy and NMR 259 experiments. 260 261 Mechanically relevant glycated lysines in I91 262 Our results so far indicate that MG-induced glycation of I91 le ads to the formation of intradomain lysine-lysine 263 crosslinking AGEs in the absence o f major structural alteration s. To directly examine the extent of this type of 264 AGEs, we did additional AFS experiments using MG-treated (I91)8, but now including a protein preparation where 265 we aimed to transform the lysine residues into inert CEL 49 ( Figures 4A,B, S10 ) . W e s t e r n b l o t ( W B ) a n a l y s i s 266 confirms that our experimental protocol results in high levels of CEL, which are considerably higher than in I91 267 incubated only with MG (Figure 4C ). AFS results show that control samples not subjected to CEL m odification 268 but incubated with MG display the expected increase in short unfolding steps after MG incubation (Figure 4D,E). 269 However, the appearance of short unfolding steps upon MG incubation is notably reduced in the case of the CEL-270 modified (I91) 8 preparation ( Figure 4D-F ). Hence, this set of experiments confirm that lysine-lysine co valent 271 crosslinking AGEs are responsible for the reduced mechanical unfolding length of glycated I91 domains. 272 Next, we investigated the specific lysine pairs that are preferential crosslinking sites in MG-treated I91. We first 273 analyzed the step sizes and associated forces resulting from me chanical unfolding of glycated I91. By fitting the 274 worm-like chain model of polymer elasticity 50 to the force versus step size distribution, we determined that the 275 reduction in contour length of the I91 domain containing crosslinking AGEs is ~16 nm (i.e. around 40 amino acids 276 trapped by the covalent crosslink considering a contour length of 0.4 nm per amino acid 51) (Figure 4G ). We 277 identified three candidate pairs of structurally vicinal lysine residues in the sequence of I91 that could account for 278 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 10 this reduction in contour length, i.e. K6-K55, K35-K79 and K37- K85 (Figure 4H). To prove that MG-induced 279 crosslinking involves at least one of these lysine pairs, we produced the mutant polyprotein (I91-KA)8, in which we 280 replaced K6, K37 and K79 with non-reactive alanine residues in all I91 domains. Upon 72 h incubation of (I91-281 KA)8 at 37ºC with 50 mM MG and subsequent purification (Figure S11), we conducted unfolding experiments by 282 AFS. Different from the wild-typ e protein, the population of 9- 13 nm steps in MG-treated (I91-KA) 8 i s o n l y 283 marginally increased with respect to the mutant protein incubated in the absence of MG ( Figure 4I,J ). In 284 combination, our results with (I91-KA)8 confirm that at least one of the selected lysine pairs is a primary target for 285 MG-induced crosslinking in I91. I n agreement with this result, following an MS-based search for crosslinked 286 peptides between lysine pairs in (I91)1 incubated with 50 mM MG for 24 h at 37°C, we were able to unambiguously 287 detect the presence of a MOLD-crosslinked dipeptide involving K35 and K79 (Figure 5). 288 289 Interdomain crosslinking AGEs targeting titin domains 290 In addition to intradomain crosslinking of lysine residues, int erdomain and intermolecular crosslinking reactions 291 should also be possible when MG reacts with serially linked protein domains such as those in (I91)8 (Figure S12A). 292 While our preparations are devoid of high-molecular weight intermolecular crosslinks thanks to the size-exclusion 293 purification step ( Figure S3B ), we cannot rule out contributions of intramolecular, interdom ain crosslinks that 294 would reduce the number of mechanically unfoldable domains in A FS experiments ( Figure S12B). Indeed, we 295 detect a slight decrease in unfolding events per AFS trace, from an average of ~5 in controls to ~4 in glycated (I91)8 296 (Figure S12C ), suggesting that MG incubation induces some interdomain cross linking. Considering that 297 interdomain crosslinks would also cause structural modifications on the (I91)8 protein, we studied the impact of MG 298 treatment on the overall structure and size of (I91)8 by small-angle X-ray scattering (SAXS). The scattering signals 299 for both glycated and non-glycated (I91)8 overlap to a great extent, suggesting that the influence of glycation on the 300 overall structure of (I91) 8 is minor ( Figure S12D). In addition, neither Porod-Debye plot shows a Porod plateau 301 (Figure S12E), proving that neither protein preparation displays compact gl obular structures and that both have a 302 certain degree of flexibility. This observation agrees with the plateau observed in the q3ꞏI(q) vs. q3 plots (Figure 303 S12F) which is typical of flexible partially folded structures 52. Moreover, both SAXS-derived pair-distance 304 distribution functions ( Pr; probability distribution of the interatomic vectors inside t he molecule) exhibit a local 305 maximum at ~20Å, which accounts for the inter-atomic distances within a single I91 domain, and another one at 306 ~46 Å, which indicates the separation between the mass centers of two consecutive I91 domains ( Figure S12G). 307 This separation agrees with the average length of the I91 domai n (Figure 3A), thus indicating extended domain 308 organizations. 309 310 In summary, AFS and SAXS data taken together point to limited extent of interdomain crosslinking in MG-treated 311 (I91)8 in our experimental conditions. 312 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 11 313 Crosslinking glycation products in full-length titin 314 Since MG-induced glycation results in the formation of crosslinking AGEs in I91, we hypothesized that other titin 315 domains could also be target of these modifications. To test th is possibility, we examined by AFS and HaloTag 316 technology53 the consequences of MG-induced glycation of native cardiac tit in isolated from mice. In these 317 experiments, we covalently attached engineered titin molecules containing a HaloTag inserted in the distal I-band 318 region of the protein to glass surfaces coated with HaloTag ligand (Figure 6A). Next, we added 50 mM MG to the 319 fluid chamber of the AFS and started pulling from titin molecules in constant-velocity mode for several hours. We 320 also ran experiments in the absence of MG. Two representative force-distance traces are shown in Figure 6A; both 321 of them exhibit a characteristic sawtooth pattern, a well-known single-molecule fingerprint in constant-velocity 322 force spectroscopy where every peak originates from the mechani cal unfolding of a titin domain and the distance 323 between peaks reflects the increase in contour length associated with unfolding54. Using the worm like chain model, 324 we find that the commonest change in contour length between con secutive peaks in the absence of MG is ~30 nm 325 (Figure 6B), as expected for non-modified titin domains53. In agreement with previous reports, we also detect a less 326 frequent population of unfolding transitions (less than 20% of total events) characterized by changes in contour 327 length shorter than 20 nm (Figure B )42,53. Interestingly, the distribution of unfolding events shifts to wards shorter 328 changes in contour length when titin is incubated with MG (Figure 6A,B), indicating that many domains in titin in 329 addition to I91 undergo MG-induced crosslinking. 330 331 Mechanical properties of glycated titin domains 332 Although intramolecular crosslinks are generally considered to stiffen proteins due to the associated reduction in 333 total contour length, this effect can be countered or even reve rsed depending on how they affect the mechanical 334 (un)folding dynamics of the targeted domains; for instance if c rosslinks result in mechanical destabilization that 335 favors unfolded polypeptides43. Hence, to quantify in detail how crosslinking AGEs affect the global mechanical 336 properties of titin domains, as well as the contribution of non -crosslinking AGEs, we further exploited our single-337 molecule force-ramp experiments. 338 339 First, we determined whether AGEs alter the mechanical stability of the I91 domain. We find that domains 340 containing MG-derived crosslinking AGEs unfold at a slightly lo wer force than counterparts incubated in the 341 absence of MG (142 and 158 pN, respectively; Figure 6C ). Similarly, CEL-containing I91 domains appear 342 mechanically weaker than the corresponding controls (114 vs 122 pN, respectively; Figure 6D). Fitting the Bell-343 Evans model of force-activated reactions55, we estimate that unfolding rates at zero force are between 3 and 8 fold 344 higher for glycated domains, while the distance to the transition state is mostly preserved indicating no major 345 changes in the force dependency of unfolding (Table S2). In combination, our analyses of unfolding forces indicate 346 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 12 that both crosslinking and non-crosslinking AGEs induce mechani cal weakening of the targeted domains, a 347 softening effect that in the case of crosslinking modifications opposes stiffening that results from reduced total 348 contour length. 349 350 Next, we did unfolding-quench-probe AFS experiments to study mechanical folding of glycated I91 (Figure 6E)29. 351 In these experiments, (I91) 8 is subjected to an unfolding force-ramp, followed by a quench pulse to 0 pN where 352 domains first collapse and subsequently regain native mechanical stability in a time-dependent manner. In the final 353 probe pulse, domains that refolded during the quench pulse unfo ld again. Folding fractions are calculated as the 354 ratio between the number of unfolding events observed in the probe pulse and those observed in the unfolding pulse. 355 To have better resolution of folding fractions at low folding t imes, we included force pulses to 40 pN before and 356 after the quench at 0 pN. Since 40 pN is non-permissive for fol ding of the I91 domain 41, the initial pulse allows 357 better synchronization of actual folding times limiting the eff ects of variable collapse ti mes. Similarly, jumping 358 straight to 40 pN avoids folding reactions occurring during the probe force ramp. Results of the unfolding-quench-359 probe AFS experiments indicate that CEL moieties do not noticeably modify the folding kinetics of the I91 domain 360 (Figure 6F). In contrast, I91 domains containing crosslinking AGEs refold at a rate ≥30 s-1, which is the resolution 361 limit of our setup implying at least a 6-fold increase compared to the corresponding control (Figure 6G). Hence, in 362 addition to stiffening resulting from reduced total contour length, crosslinking AGEs entail an additional stiffening 363 effect as a consequence of increased folding of targeted domains. 364 Prevailing stiffening induced by titin glycation 365 As explained in the previous section, our single-molecule data demonstrate that, similar to oxidative modifications, 366 glycation of titin can entail both stiffening and softening effects29,42. To illustrate the available range of modulation 367 of titin mechanics by AGEs, we did Monte Carlo computer simulations as reported 43. We first built virtual models 368 of the I-band of human titin containing 104 (for the N2BA isoform) or 48 (for the N2B isoform) Ig domains, all of 369 them with equivalent mechanical properties, as well as random c oil N2Bus and PEVK regions ( Table S3). Since 370 the PEVK region is rich in lysine residues ( Figure 6H,I ), we also estimated the reduction in contour length 371 associated with crosslinking modifications targeting this region of titin. Specifically, considering that the first and 372 last lysine in each of the 31 PEVK repetitions can form a crosslink (Figure 6I), we used a 67% maximum reduction 373 in contour length upon glycation of the PEVK region (Figure 6J ). We acknowledge that crosslinks between 374 different PEVK repetitions could result in further decreases in contour length; however, we have not contemplated 375 this possibility here. 376 In the simulations, we subjected the virtual I-bands to 1-Hz tr iangular force pulses between 0 pN and a predefined 377 setpoint peak force while monitoring the resulting changes in t itin length ( Figure 6K,L ). During these 378 extension/relaxation cycles, titin domains unfold and refold stochastically according to their folding and unfolding 379 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 13 rates, which depend on the type of glycation modification as de tected in our single-molecule experiments (Tables 380 S4, S5). At t = 0 s, all domains are folded and, as the simulations p rogress, a fraction of domains transitions to the 381 unfolded state resulting in longer protein lengths (Figure 6L). Simulation times were long enough to always reach 382 steady-state lengths in all simulation runs (Figures 6M, S13). 383 In Figure 6M, we present the results of the simulations of N2BA titin at a peak force of 10 pN, typically considered 384 the upper limit of the physiological range 40, comparing control conditions with two extreme scenarios where all 385 domains are glycated with either non-crosslinking or crosslinking AGEs. These results readily capture the prevailing 386 stiffening effect of crosslinking modifications, which cause ~33% shortening of titin at peak force. In contrast, non-387 crosslinking modifications result in subtle, yet measurable, softening. Prompted by these results, we ran additional 388 simulations at intermediate degrees of glycation and considerin g different proportions of non-crosslinking and 389 crosslinking AGEs. In Figure 6N, we compare steady-state lengths of titin in these simulations with respect to 390 control simulations. As expected, we find that the mechanical effects of both types of AGEs progressively increase 391 at higher glycation fractions. More importantly, we observe tha t stiffening contributions of crosslinking AGEs 392 outweigh softening induced by non-crosslinking counterparts even when the ratio of crosslinking/non-crosslinking 393 modifications is only 20/80, independently of the total levels of glycation (Figure 6N). We observe similar effects 394 for the N2B isoform, although in this case stiffening is less apparent ( Figure S13 A,B). Increasing the peak force 395 of our simulations to 100 pN exacerbates the mechanical consequences of both crosslinking and non-crosslinking 396 modifications, with up to 57% reduction in length induced by crosslinking AGEs (Figure S13C-F). Further analysis 397 of the Monte Carlo simulations captures a fundamental role of g lycation of the PEVK domains in global titin 398 stiffening (Figure S14). Indeed, if we consider that no glycation targets the PEVK re gion, no stiffening of titin is 399 detected at 10 pN peak force indicating no overall mechanical effect of crosslinking AGEs targeting Ig domains at 400 this force (Figure S14A,C). At 100 pN in the absence of PEVK glycation, the stiffening contribution of crosslinking 401 AGEs in Ig domains becomes apparent again (Figure S14B,D). Such force dependency of the stiffening effects of 402 crosslinking AGEs in titin Ig domains resembles the behavior reported for intradomain disulfide bonds43. 403 Finally, we also quantified from our Monte Carlo simulations how glycation affects Ig domain unfolding/refolding 404 dynamics, which can influence downstream force-dependent interactions and PTMs11,15 and active force generation 405 by sarcomeres41. With this aim, we counted the number of unfolding events in t he different simulations. Results 406 indicate that both crosslinking and non-crosslinking AGEs increase the extent of Ig domain (un)folding transitions, 407 particularly at low forces (Figure S15). 408

409 High concentrations of carbohydrates and the associated toxic s ide-products of glycolysis (e.g. -oxoaldehydes) 410 are at the basis of many human diseases including diabetes, obe sity, age-related disorders and the metabolic 411 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 14 syndrome, as well as their associated comorbidities (e.g. cardiovascular complications)56. However, the underlying 412 molecular pathomechanisms are yet to be fully elucidated. A mai n example is the origin of myocardial stiffening 413 typically associated with carbohydrate-related diseases, a majo r risk factor for developing heart failure 3. While 414 fibrosis and glycation of ECM components have been shown to stiffen the myocardium in these conditions6,7,27, the 415 reasons for the concurrent rigidification of cardiomyocytes (ev en permeabilized ones) remain incompletely 416 understood, limiting our grasp on how diastolic failure develop s in affected patients. So far, stiffening of diabetic 417 cardiomyocytes has been proposed to be contributed by hypophosphorylation of titin and by glycogen accumulation, 418 although compensatory overexpre ssion of the compliant N2BA titi n isoform complicates quantitative 419 interpretation6,33,57,58. The data presented here uncover an additional stiffening effe ct induced by titin glycation, a 420 modification found in the myocardium of both animal models and humans in diabetes34 (Figure 1C-D) and aging35. 421 Specifically, our work demonstrates that MG-induced glycation, a main chemical route during glycative stress 20, 422

in cardiomyocyte stiffening ( Figure 1E-I) because it leads to the preferential formation of crosslinkin g 423 AGEs within titin domains. Indeed, Monte Carlo simulations indi cate that stiffening induced by intradomain 424 crosslinking AGEs prevails over softening contributions by non- crosslinking counterparts already when the 425 proportion of crosslink-containing domains reaches 20% (for N2BA titin) or 40% (for N2B titin), independently of 426 the total extent of glycation ( Figures 6N, S13 ). From our single-molecule experiments at saturating glycation 427 conditions, we estimate the actual proportion of crosslink-cont aining domains to be at least close to 80% ( Figures 428 2N, S4), which agrees with the observation using WB that the percentage of CEL within all the MG-derived AGEs 429 is relatively low (Figure 4C). This level of crosslinking modifications consistently leads to titin stiffening in Monte 430 Carlo simulations, in line with our cell mechanics experiments. Since simulations also show that glycation of the 431 PEVK region is a main contributor to titin stiffening (Figure S14), the different proportions of crosslink-containing 432 domains required to achieve stiffening in N2BA and N2B titin ca n be easily explained by the distinct relative 433 mechanical contribution of the PEVK domains in both isoforms (the ratio PEVK length in nm/number of Ig domains 434 is 6.9 and 1.4 for the N2BA and the N2B isoforms, respectively) ( Table S3). To strengthen our conclusions, we 435 have obtained further evidence of the presence of crosslinking AGEs in MG-treated titin domains by mass 436 spectrometry (Figure 5) and NMR ( Figure 3C). Furthermore, our NMR results show that the structure of I91 is 437 mostly unaffected by MG-induced glycation ( Figures 3D, S7 ), which induces instead restricted conformational 438 entropy of the protein as a result of the formation of crosslin king AGEs ( Figures 3F-H, S8A, S9 ). Such overall 439 preservation of the I91 native fold is also supported by the mo dest reduction in mechanical stability of glycated 440 domains (Figure 6C) and their enhanced/preserved ability to refold ( Figure 6F,G). Hence, in combination, our 441 findings add to available evidence challenging the conventional view that glycation universally compromises 442 protein structure17. 443 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 15 A potential limitation of our work stems from the millimolar MG concentration we used to trigger glycation in 444 timescales compatible with single-molecule and cell mechanics e xperiments. This concentration is several orders 445 of magnitude above typical estimates in cells, which indicate t hat MG levels do not exceed the micromolar range 446 even in situations of glycative stress 59. Hence, it appears that the increased glycation of titin in di abetes and aging 447 that we and others have observed ( Figure 1C,D)34,35 most probably results from the slow accumulation of AGEs 448 involving multiple glycation pathways, a situation challenging to reproduce in the laboratory. Still, we have been 449 able to detect a measurable fraction of crosslinking AGEs occur ring very rapidly after addition of 50 mM MG 450 (Figure 2N ). This observation points to favored glycation reactions induced by MG that can proceed at 451 physiological concentrations. It is also important to consider that given the high reactivity of MG and that the 452 glycolytic flux can be estimated to produce ~125 mol of MG per kg of cell mass per day (i.e. 133 M per day for 453 a 15,000 m3 cardiomyocyte considering a cellular density of 1.06 g/mL)20, it is conceivable that the effects of any 454 given steady-state concentration of MG in vivo are noticeably higher than those of a seemingly equivalent, but not 455 replenishable, concentration of MG in in vitro experiments. 456 Several aspects of our work call for further research. First, given that intracellular protein glycation is common to 457 many tissues in addition to the myocardium 60, we speculate that equivalent mechanisms to the ones we descri be 458 here can contribute to stiffening of cell types other than card iomyocytes. Similarly, it is possible that glycation 459 targeting protein-folding-based mechanosensor proteins like talin can have profound consequences in cell 460 behavior61. In this regard, although the existence of crosslinking AGEs can be extrapolated from the presence of 461 non-crosslinking counterparts19, new methods are required to quantify crosslinking modifications in native proteins 462 and how they depend on cellular state. These new methods will need to overcome both the vast combinatorial space 463 of potential target sites, particularly in large proteins like titin, and the interpretation and scoring of hybrid 464 fragmentation spectra, which is very challenging for convention al mass-spectrometry approaches 62,63. For further 465 quantitative understanding of the effects of protein AGEs in ce ll mechanics, additional work will need to address 466 protein folding intermediates41 and to integrate both stiffening interdomain crosslinks and the effects of mechanical 467 crosstalk with other PTMs, especially those competing for glycation sites like acetylation and ubiquitination64. 468 Concluding remarks 469 Our results with titin add to the list of detrimental functional consequences that follow glycation of key proteins in 470 cardiomyocytes64,65, including other sarcomere components important for contraction 34,66-68, the calcium pump 471 SERCA2a69 and the calcium channel RyR235. At the methodological level, our work illustrates how single-molecule 472 AFS can discriminate the mechanical consequences of crosslinking and non-crosslinking AGEs in a manner that is 473 chemically agnostic, therefore overcoming intrinsic heterogenei ty of non-enzymatic glycation reactions. Having 474 pinpointed crosslinking modifications as culprits of cell rigid ification, our work provides deeper understanding of 475 highly promising therapeutic strategies based on scavenger molecules that block protein glycation 27. According to 476 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 16 our data, further improvements could be achieved by preferentially preventing the formation of crosslinking AGEs, 477 or even by inducing their cleavage. Single-molecule force spectroscopy methods can serve as a valuable screening 478 platform to this aim. 479 480 Author contributions: 481 Conceptualization: A.B., A.C.D., E.H.G., M.A., Jor.A.C. 482 Methodology: A.C.M., E.C., R.S.R., N.V., D.S.O., I.M.M., D.V.C. , M.R.P., F.M.E., Joa.A.C., C.P.M., R.G., J.V., 483 I.F.P. 484 Investigation: A.B., A.C.D., A.C.M., E.C., E.H.G., M.A., Jor.A.C. 485 Funding acquisition: J.V., Jor.A.C. 486 Writing – original draft: A.B., A.C.D., M.A., Jor.A.C. 487 Writing – review & editing: all authors 488 489 Competing interests: Authors declare that they have no competing interests. 490 491 Data availability 492 The authors declare that the data supporting the findings of th is study are available within the paper and its 493 Supplementary Information files. Should any raw data files be needed in another format they are available from the 494 corresponding authors upon reasonable request. The NMR assignments have been deposited in the Biological 495 Magnetic Resonance Data Bank (BMRB) under the accession codes 5 3389 and 53390 for (I91) 1 and MG-treated 496 (I91)1, respectively. 497 498 Code availability 499 The custom-built code used in this study is available from the corresponding authors upon reasonable request. 500 501

502 Jor.A.C. acknowledges funding from Ministerio de Ciencia, Innov ación y Universidades (MICIU, 503 MICIU/AEI/10.13039/501100011033) through grants PID2020-120426G B-I00, PID2023-147683NB-I00, 504 PLEC2022-009235 and RED2022-134242-T, the Regional Government o f Madrid (grant Tec4Bio S2018/NMT-505 4443, 50% co-financed by the European Social Fund and the Europ ean Regional Development Fund for the 506 programming period 2014-2020, and grant TecNanoBio TEC-2024/TEC -158, Bases Reguladoras 2402/2024; 507 Convocatoria 3177/2024), and the European Research Council (ERC ) under the European Union’s Horizon 2020 508 research and innovation program (grant agreement No. [101002927 ]). J.V. acknowledges grants PID2021-509 122348NB-I00 and PID2021-126827 OB-I00 funded by MICIU and by “E RDF A way of making Europe”, 510 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 17 PLEC2022-009298, PLEC2022-009235 and EQC2021-007053-P funded by MICIU and by “European Union 511 NextGenerationEU/ PRTR”, and S2022/BMD-7333-CM (INMUNOVAR-CM) funded by Comunidad de Madrid. 512 The project leading to these results has received funding from ‘la Caixa’ Foundation under the project code 513 LCF/PR/HR22/52420019. CNIC is supported by the Instituto de Sal ud Carlos III (ISCIII), the MICIU and the Pro 514 CNIC Foundation, and is a Severo Ochoa Center of Excellence (gr ant CEX2020-001041-S funded by MICIU). 515 A.B. was the recipient of an FPI predoctoral fellowship (PRE201 8-084392 funded by MICIU). A.C.D. 516 acknowledges funding from ‘la Caix a’ Foundation (LCF/BQ/PI22/11 910029) and the MICIU through project 517 PID2022-140352NA-I00. A.C.D. is a Research Associate of the Bel gian FNRS. A.C.M. is a recipient of a 518 predoctoral fellowship funded by the Regional Government of Mad rid (PIPF-2023SAL-GL-31131). R.S.R. 519 acknowledges funding from the European Molecular Biology Organi zation (EMBO, postdoctoral fellowship 520 EMBO ALTF 417-2022). I.M.M. was the recipient of a Doctoral INPhINIT fellowship from ‘la Caixa’ Foundation 521 (ID100010434, fellowship code LCF/BQ/DR20/11790009). I.F.P. ack nowledges funding from Fundação para a 522 Ciência e Tecnologia (2023.17125.ICDT). We thank the personnel from CNIC animal housing, Matthew M. 523 Borkowski (Aurora Scientific) for his help with the Permeabilized Myocyte Test System 1600A, Guadalupe Sabio 524 and María Villalba for their advice on animal models, Jaime Andrés Rivas-Pardo for advice on titin purification 525 and Jonathan A. Kirk and Álvaro Martínez-del-Pozo for feedback. We thank Gabriel Martorell and Rosa Gomila, 526 both from the Scientific and Technical Services of the University of the Balearic Islands, for their assistance in 527 performing NMR and MALDI-TOF/TOF experiments. We thank all memb ers of the Molecular Mechanics of the 528 Cardiovascular System team for their support and input. The authors thank the donors and the Hospital Universitario 529 Puerta de Hierro Majadahonda (HUP HM)/Instituto de Investigación Sanitaria Puerta de Hierro-Segovia de Arana 530 (IDIPHISA) Biobank (Carlos III Health Institute Biomodels and B iobanks Platform – PT23/00015) for the human 531 specimens used in this study. 532 533

534 Human subject research 535 Samples and data from patients included in this study (Table S6) were provided by the Hospital Universitario Puerta 536 de Hierro Majadahonda (HUPHM)/Instituto de Investigación Sanita ria Puerta de Hierro-Segovia de Arana 537 (IDIPHISA) Biobank (Carlos III Health Institute Biomodels and Biobanks Platform – PT23/00015). Samples were 538 processed following standard operating procedures with the appropriate approval of the Ethics and Scientific 539 Committee (Ref: 67/187906.9/24). The handling of samples and patient data was conducted in accordance with the 540 principles of the Declaration of Helsinki and the International Conference on Harmonization of Good Clinical 541 Practice guidelines, under the framework of Spanish and Europea n regulations concerning data protection and 542 clinical research. 543 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 18 Animal research 544 Mice were housed and maintained in CNIC’s animal facility in accordance with Spanish and European Legislation 545 (Directive 2010/63/EU amended by Regulation EU 2019/1010; CNIC-01/18- PROEX042/1; CNIC-01/23 - PROEX 546 107.8/23). Only male mice were used for the experiments. Adult mice were sacrificed using CO 2, and neonates 547 were sacrificed by decapitation. Hearts were isolated, perfused with PBS, frozen in liquid nitrogen, and stored at -548 80ºC. All procedures on rats were reviewed and approved by the Faculty of Medicine of the University of Porto 549 (FMUP) Animal Welfare and Ethics Review Body (Órgão Responsável pelo Bem-Estar dos Animais [ORBEA-550 FMUP]) and the Portuguese competent authority (Direção Geral de Alimentação e Veterinária [DGAV], reference 551 number 8161/23-S) and performed in accordance with EU Directive 2010/63/EU and Decreto-lei 113/2013 national 552 legislation at the FMUP animal facility. Seven-week-old male Wi star-Han rats were acqu ired from Charles River 553 Laboratories (Barcelona, Spain) and housed in groups of 4 anima ls per cage in a controlled environment under a 554 12:12-h light-dark cycle at a RT of 22ºC, with free supply of f ood and water. Twelve weeks later, animals were 555 anesthetized with sevoflurane (8%), the heart was removed and immediately frozen in liquid nitrogen for subsequent 556 analysis. All procedures were carried out by properly trained and licensed researchers. 557 558 559 LC-MS/MS mass spectrometry in native titin experiments 560 Myocardial protein extracts in the presence of N-ethylmaleimide were obtained following Herrero-Galán et al. 43. 561 Samples were run in 3.5% SDS-PAGE gels in the absence of reduci ng agents. Proteins were visualized with 562 Coomassie blue and titin bands were sliced out from the gel and stored at 4ºC until analysis. Bands were equilibrated 563 in 50 mM ammonium bicarbonate (ABC) prior to reduction with 50 mM DTT and alkylation with 100 mM 564 iodoacetamide, both in 100 mM ABC. Gel bands were then subjecte d to in-gel-digestion using modified bovine 565 chymotrypsin, sequencing grade (Promega) at a final ratio of 1:20 (chymotrypsin-protein). Digestion proceeded 566 overnight at 37ºC in 100 mM ABC, pH 7.8. After digestion, pepti des were extracted using acetonitrile with 0.1% 567 (v/v) trifluoroacetic acid (TFA). Finally, peptides were desalt ed and dried until LC-MS analysis. LC-MS analysis 568 was done using an Evosep One HPLC (Evosep) coupled to an Orbitrap Eclipse Tribrid Mass Spectrometer (Thermo 569 Fisher Scientific) using an Endurance Evosep column 15 cm x 150 m ID as analytical column (Thermo Fisher 570 Scientific) coupled to a stainless steel emitter of 30 m ID. Peptides were eluted from Evotips and analyzed using 571 the Evosep One pre-programmed gradient for 15 samples per day (SPD). MS spectra were acquired in the Orbitrap 572 analyzer using full ion-scan mode with a 390-1700 m/z range and 60,000 FT resolution. The automatic gain control 573 target was set at 1 x 10 6 with 50 ms maximum injection time. HCD fragmentation was perfo rmed at 30% of 574 normalized collision energy and MS/MS spectra were analyzed at a 30,000 resolution in the Orbitrap with automatic 575 gain control target set at 1 x 10 5 and 100 ms maximum injection time. For peptide identification, tandem mass 576 spectra were extracted, and charge state was deconvoluted by Proteome Discoverer 2.5.0.400 (Thermo Fisher 577 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 19 Scientific). MS/MS spectra were analyzed using SEQUEST HT (Thermo Fisher Scientific) performing searches 578 against a titin database43 considering as variable modifications methionine oxidation (mass = 15.995), NEM and 579 IAM-modified cysteines ( mass = 125.048 and 57.021, respectively), and the masses of non -crosslinking AGEs 580 CML (mass = 58.005), CEL (mass = 72.021) and MG-H ( mass = 54.011). Peptide-spectrum matches (PSM) 581 were filtered to a q-value < 0.01. 582 583 Purification of methylglyoxal 584 MG was purchased as a 40% solution (Sigma-Aldrich) and addition ally purified by steam distillation. The 585 concentration of MG after purification was quantified by incuba ting with excess H 2O2 and titrating the remaining 586 H2O2 with KMnO4 according to Friedemann’s method70. The fractions were frozen until use. 587 588 AFM-based nanomechanical spectroscopy on neonatal cardiomyocytes 589 Dissected hearts from neonatal (postnatal days 1-3) mice were m inced with scissors in cold Hanks’ Balanced Salt 590 Solution (HBSS) and dissociated using the Pierce Primary Cardiomyocyte Isolation Kit (Thermo Fisher Scientific) 591 in 0.21 mL enzyme mix/heart in a 2 mL reaction tube for 20 min at 37°C under constant end-to-end rotation. The 592 pellet was centrifuged, washed twice with HBSS and resuspended in 0.5 mL cardiomyocyte medium (DMEM for 593 Primary Cell Isolation containing 10% heat-inactivated FBS and 1% penicillin/streptomycin)/heart and plated onto 594 MatTek dishes covered with matrigel solution (Fisher Scientific ). AFM measurements were performed with a 595 commercial JPK NanoWizard III inst rument (Bruker-JPK) coupled t o an inverted Axio Observer A1 optical 596 microscope (Carl Zeiss). Cardiomyocytes were permeabilized with 0.2% (v/v) Triton X-100 in 1% (w/v) BSA for 597 15 min and incubated with 50 mM MG or PBS for 4 h at 37ºC. Foll owing rinsing with PBS, cardiomyocytes were 598 probed at RT in PBS using rectangular Si 3N4 AFM cantilevers with silicon tips of radius < 15 nm and a nomi nal 599 spring constant of 0.1 N m -1 (BioLever mini, Bruker). Cantilevers were calibrated using the thermal fluctuation 600 method71. FD curves were recorded in contact mode on the surface of single cardiomyocytes to determine Young’s 601 moduli. Approach-retract cycles were performed with a tip velocity of 10 µm s-1, a ramp size of 2 µm and a setpoint 602 force of 2 nN. We analyzed around 100 FD curves per cardiomyocyte. Young’s modulus values (E) were calculated 603 using the JPK Data Analysis software, by fitting the approach section of the curve to the Hertz model using Equation 604 172: 605 606 𝐹ൌ ସ ଷ ா ଵିఔమ√𝑅𝛿ଷ (1) 607 608 In Equation 1, F is the applied force, R is the probe radius, ν is Poisson’s ratio, and δ is the indentation. Cells mostly 609 consist of water and are considered incompressible, hence a Poisson ratio of 0.5 was used. 610 611 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 20 Tensile testing of adult rat cardiomyocytes 612 Rat cardiomyocytes were isolated from cardiac tissue of 19-week -old male Wistar rats. Cells were isolated in 613 imidazole-containing relaxing buffer (RB, Table S7) by homogenizing a piece of tissue from the left ventricle with 614 ceramic microspheres (Bertin Instruments) for 10 seconds at 5,0 00 rpm using a MiniLys system (Bertin 615 Instruments). The cell suspension was resuspended to a final vo lume of 2.5 mL. Cells were incubated with 0.5% 616 Triton X-100 (Millipore) for 5 min on ice. Subsequently, the detergent was removed by several washes in 15 mL 617 imidazole-containing RB followed by centrifugation of the cells for 1 min at 348 g. Passive tension measurements 618 were performed using a Permeabilized Myocyte Test System 1600A (Aurora Scientific), where single isolated 619 skinned cardiomyocytes were fixed to a force transducer and a motor. Cells were transferred to propionic-containing 620 RB (Table S7) and passive force was measured during 6 step length increment s of 5% L0. To avoid quenching of 621 MG with phosphocreatine, cells were incubated with MG (or contr ols in the absence of MG) for 30 min at in 622 phosphocreatine-free RB (Table S7). Finally, the cells were transferred back to propionic-contai ning RB and the 623 passive tension was measured again. All force values were converted to tension values assuming an elliptical shape 624 of the cell, according to 625 626 𝐶𝐴 ൌ 0.7 𝜋 ௔మ ସ (2) 627 628 where a is the width of the cardiomyocyte observed in the microscope and CA is the estimated cross-sectional 629 area73. In control experiments to check for the potential effect of A TP modification by MG, passive tension of 630 cardiomyocytes was first measured in phosphocreatine-free RB. Next, cells were incubated for 30 min in a 631 phosphocreatine-free RB that had been previously incubated for 30 min with 50 mM MG (excess MG in the buffer 632 was quenched with 0.1 M Tris-HCl pH 7.5; ATP-modified RB), and passive tension was determined. To account for 633 non-specific effects of adducts between Tris and MG, we included a control experiment where Tris was added to 634 MG before final addition to phosphocreatine-free RB (ATP-preserved RB). 635 636 Protein expression and purification 637 The complementary DNA (cDNA) coding for monomers and octamers o f titin I91 was cloned in the pQE80 638 expression plasmid (Qiagen). The cDNA coding for titin (I91-K6/37/79A)8 synthetized by GeneArt Gene Synthesis 639 (Thermo Fisher Scientific) was inserted in the pQE80 expression p l a s m i d u s i n g B a m H I a n d K p n I r e s t r i c t i o n 640 enzymes and verified by Sanger sequencing. Polyproteins were ex pressed in Escherichia coli B L R ( D E 3 ) . 641 Monomeric protein was expressed in Escherichia coli BL21 (DE3) and cultured in isotopically labeled M9 minimal 642 medium (Table S8) containing 13C6-D-glucose and 15N-ammonium chloride. Fresh cultures (OD600 = 0.6-1.0) were 643 induced with 1 mM isopropyl -D-1-thiogalactopyranoside (IPTG) for 3 h at 37ºC and 250 rpm agitation. Bacteria 644 culture was lysed using French Pr ess. Proteins were purified by affinity chromatography using gravity-based 645 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 21 strategy or in a HisTrap Fast Flow column using Fast Protein Liquid Chromatograp hy (FPLC) AKTA Pure 25 L 646 system (GE Healthcare). Proteins were eluted in 50 mM sodium phosphate pH 7, 300 mM NaCl, 250 mM imidazole, 647 followed by protein concentration using Amicon Ultracel 3K, for (I91) 1, or 30K, for (I91) 8, centrifugal filter 648 (Millipore). Finally, we did an additional purification step us ing size-exclusion chromatography in Superdex 200 649 Increase 10/300 GL or Superose 6 Increase 10/300 GL columns in the AKTA Pure 25 L system in Hepes (10 mM 650 Hepes, 150 mM NaCl, 1 mM EDTA, pH 7.2) or in 0.2 M phosphate (p H 7.4). Purity of samples was checked by 651 SDS-PAGE. Protein concentration determinations considered theoretical extinction coefficients values (Table S9). 652 653 Protein glycation 654 For batch glycation, Ni-NTA fract ions with the highest protein concentration were pulled together, followed by 655 buffer exchange to 0.2 M phosphate pH 7.4 using PD-10 desalting columns (GE Healthcare) or by dialysis using 656 Slide-A-Lyzer Dialysis Casettes (Thermo Scientfic) with 3,500 D a molecular weight cut-off (MWCO) for (I91) 1 657 and 20,000 Da MWCO for (I91)8 proteins for 24 h at 4ºC, 100 rpm agitation. Reactions with MG were done at 37ºC 658 and excess MG was removed by size-exclusion chromatography usin g FPLC as described above. To induce 659 formation of CEL, the protein was incubated with 70 mM pyruvic acid (Sigma-Aldrich) and 100 mM sodium 660 cyanoborohydrate (NaBH3CN) (ThermoFisher Scientific) for 48 h at 50ºC 49, and excess reagents were removed as 661 above. For proteins treated both with pyruvic acid and MG, a di alysis step in 0.2 M phosphate buffer pH 7.4 was 662 included between both reaction steps. 663 664 MALDI-TOF/TOF mass spectrometry 665 Protein preparations were dialyzed against mili-Q water. For (I91)1, 2 μl of the protein solution were mixed in a 1:1 666 ratio with a matrix solution containing 5 mg/mL of α-cyano-4-hydroxycinnamic acid prepared in a water:acetonitrile 667 (70:30). For (I91) 8, the protein solution was mixed in a 1:1 ratio with a matrix s olution containing 10 mg/mL of 668 sinapinic acid prepared in a water:acetonitrile (50:50). Both matrix solutions contained 0.1% TFA. 0.5 μL aliquots 669 of those mixtures were spotted onto a steel target plate (MTP 384), air-dried, and subjected to mass determination. 670 Mass spectra were acquired on a Br uker Autoflex III MALDI-TOF/T OF spectrometer equipped with a 200-Hz 671 smart-beam pulsed N2 laser (λ 337 nm). The IS1 and IS2 voltages were 20 kV and 18.4 0 kV respectively, and the 672 lens voltage was 8.4 kV. Measurements were performed using a po sitive reflector mode with matrix suppression 673 below 4000 Da. External calibration was performed using a standard protein mixture. 674 675 Single-molecule force spectroscopy by Atomic Force Microscopy 676 Single-molecule AFS measurements were done in a Luigs & Neumann setup following published protocols 44. 677 Briefly, 5-10 L of purified protein solution were spread on 15-mm-diameter coverslips coated with 100-nm-thick 678 gold (Luigs & Neumann). Experiments were performed in 0.2 M pho sphate buffer pH 7.4 using silicon nitride 679 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 22 MLCT-FB cantilevers (Bruker), coated with 60 nm titanium-gold on their back side. Cantilevers were calibrated by 680 the thermal noise method71. Single polyproteins were picked up by pressing the tip against the coverslip for 1 second 681 at 1,500 pN contact force. To trigger protein unfolding, a 40 pN s-1 force ramp for 6.5 seconds was applied. For data 682 selection and analysis, two levels of fingerprinting were used to select single-molecule events. As a non-stringent 683 fingerprint, all the traces containing at least one step of ~25 nm, which corresponds to one I91 unfolding event 29, 684 were selected. The step size of all events appearing on the sel ected traces were measured and plotted on 685 bidimensional histograms. Based on the results, a more stringen t fingerprint criterion, by which only traces 686 containing at least 2 events of 23-27 nm and/or 9-14 nm steps were selected, was used for further analysis. For 687 analysis of mechanical unfolding, only traces presenting detachment at forces higher than 240 pN were selected to 688 ensure full unfolding of the polyprotein. Traces presenting nonspecific events at forces >50 pN were discarded. 689 Traces with more than 8 events were only considered for analysi s in Figure S12C. To quantify the fraction of 690 glycated protein domains, we divi ded the number of short steps by the total number of steps. For quantifying 691 mechanical stability, we fit the AFS data to the Bell-Evans mod el, which considers that the mechanical unfolding 692 rate of the protein (r) depends on the applied force (F) according to: 693 694 𝑟ൌ𝑟 ଴ e ಷ∆ೣ ೖಳ೅ (3) 695 696 where r0 indicates the unfolding rate at zero force, Δ x is the distance to the transition state of the mechanical 697 unfolding reaction, kB is the Boltzmann constant and T is the absolute temperature 74. Specifically, we used the 698 following equation, which is the derivation of the Bell-Evans model for the case of a force ramp55: 699 700 𝑃௨ሺ𝐹ሻൌ1െ𝑒 𝑥 𝑝൤ ௥బ௞் ఈ∆௫∙ ൬𝑒 ಷ∆ೣ ೖ೅െ1൰൨ (4) 701 702 For folding analysis, a three-pulse experiment was used. Upon unfolding of domains in a 40 pN s-1 ramp, the force 703 was quenched to 0 pN to allow for protein collapse and refoldin g. Finally, proteins were stretched again to detect 704 domains that refolded during the quench time29. Folding fractions were calculated as the ratio between the number 705 of unfolded domains during the probe pulse and the number of un folded domains during the unfolding pulse. For 706 this analysis, we only considered traces in which the difference in protein length between first and second ramp was 707 not bigger than 15 nm. The folding ratio was quantified for varying quench times to quantify the folding rate. Errors 708 were estimated using bootstrapping75. 709 710 NMR spectroscopy 711 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 23 To prepare samples for NMR spectroscopy, 13C,15N-(I91)1 was incubated in the presence ( glycated sample) or 712 absence (non-glycated sample) of 50 mM MG for 48 h at 37ºC. Solutions containing 13C,15N-(I91)1 (270 μM) or 713 MG-treated 13C,15N-(I91)1 (200 μM) were prepared in 20 mM sodium acetate, pH 4.5 supplemented with 10% (v/v) 714 D2O. In addition, the same buffer was used to prepare a solution containing non-labeled (I91)1 (120 μM) glycated 715 with 13C-MG using the same conditions as above, and a solution contain ing 6 mM 1,3-bis((S)-5-amino-5-716 carboxypentyl)-5-methyl-1H-imidazol-3-ium acetic acid salt (sMOLD; Iris Biotech). These solutions were then 717 used for NMR studies, which were carried out at 25ºC on a Bruker Avance III spectrometer operating at a 1H 718 resonance frequency of 600.1 MHz (14.1 T) and equipped with a 5 mm 1H/13C/D-BB z-GRD triple resonance broad 719 band probe (TBI). In all experiments, water suppression was achieved by the watergate pulse sequence76 and proton 720 chemical shifts were referenced to the water signal fixed at 4.771ppm. 13C and 15N chemical shifts were referenced 721 i n d i r e c t l y u s i n g t h e 1H,X frequency ratios of the zero point 77. The spectra were processed using the software 722 packages NMRPipe/NMRDraw 78 and Topspin (Bruker), whereas the data were analyzed using Xea sy/Cara79, 723 Sparky80 and Protein Dynamics Center software (Bruker). The sequence-sp ecific backbone assignments of native 724 and MG-treated (I91)1, as well as the assignments of their C , were achieved using the following standard 2D and 725 3D NMR experiments: 1H,15N-HSQC, 15N-TOCSY-HSQC, 15N-NOESY-HSQC (250 ms), HNCACB, 726 CACB(CO)HN, HNCO, HCCH-TOCSY, HNHA, 13C-NOESY-HSQC and HN(CA)CO. To assign (I91) 1 w e 727 initially started from part of the assignment previously publis hed by Improta et al. at 35ºC 81, which we adjusted, 728 confirmed and expanded. The chemical shift assignment of MG-tre ated (I91) 1 was achieved using the NMR 729 experiments mentioned above. The NMR assignments have been depo sited in the Biological Magnetic Resonance 730 Data Bank (BMRB) under the accession codes 53389 and 53390 for (I91)1 and MG-treated (I91) 1, respectively. 731 Protein backbone assignments were used to estimate the secondar y structure content at the residue level. This was 732 carried out using three different algorithms: i) the neighbor c orrected structure propensity calculator (ncSPC) 82, 733 which bases its calculation on the random coil library and adds an additional weighting procedure that accounts for 734 the backbone conformational sensitivity of each amino acid type ; ii) the TALOS+ program 83, which uses the 735 chemical shifts and the sequence information to make quantitati ve predictions of the secondary structural content; 736 and iii) the CSI 3.0 web server, which uses backbone chemical s hifts to identify up to eleven different types of 737 secondary structures84. We also acquired 15N longitudinal (R1) and transverse (R2) relaxation data, as well as steady-738 state 15N HET-NOE data. The R1 values were determined using a series of 11 experiments with r elaxation delays 739 ranging from 10 to 2000 ms. The R2 values were obtained using a set of 11 different relaxation delays ranging from 740 2 to 140 ms. Recycle delays were 3 s in both R1 a n d R2 experiments. The 15N HET-NOE measurements were 741 performed by 3 s high power pulse train saturation with a 5 s r ecycle delay. R1, R2 and 15N HET-NOE data were 742 acquired using standard pulse sequences 85. We acquired 32 scans in R1 and R2 experiments, and 200 scans in 15N 743 HET-NOE experiment. R1 and R2 relaxation data were fitted to a mono-exponential decay functi on, whereas 15N 744 HET-NOE data were obtained as the ratio of the peak intensities from the saturated and unsaturated spectra. 745 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 24 Relaxation constants and experimental errors were calculated using the Protein Dynamics Center software (Bruker). 746 R1 and R2 relaxation constants and their experimental errors were used to determine the correlation times (c) through 747 the TENSOR2 software 86. The orientation and the magnitude of the rotational diffusion tensor were determined 748 from the R1 and R2 relaxation constants using the Quadratic-Diffusion software 87. The experimental relaxation 749 constants and the solution structure of I91 (PDB: 1TIT)81, which was modified adding the three last residues of our 750 construct (i.e. R90-C92; Text S1 using Pymol88), were studied using the molecular diffusion derived by Woessner 751 in combination with the Lipari–Szabo model-free analysis of loc al flexibility47. The model-free order parameters 752 (O2) report on the amplitudes of conformational fluctuations on ti me scales faster than the overall rotational 753 diffusion. The amide bond length was fixed at 1.02 Å, whereas the chemical shift anisotropy was fixed at -172 ppm 754 for the 15N backbone spins 89. TENSOR2 software was also used to test five different models to characterize the 755 internal dynamics of the amide groups: model 1 ( O2), model 2 (O2, e), model 3 (O2, kex), model 4 (O2, e, kex) and 756 model 5 (Of2, Os2, e)90. e is the effective internal correlation time (describes motions on a timescale > 20 ps); kex is 757 a chemical exchange term (describes slow timescale motions on the order of μs–ms); and Of2 and Os2 are terms that 758

from splitting the generalized order parameter ( O2) into two order parameters reflecting slower and faster 759 motions, respectively. The con fidence levels were estimated usi ng 100 Monte Carlo simulations per run in 760 combination with c2 and F-test criteria. The residues that had an overlapped cross-peak with other residues or they 761 were too broadened to allow the quantitative analysis were excl uded. The generalized order parameter ( O2) was 762 then used to estimate the change in the backbone conformational e n t r o p y ( ΔSconf) of (I91) 1 as a result of its 763 modification with MG. This approach is based on the assumption that the fluctuations of individual residues are 764 uncorrelated91, but it is also assumed that it underestimates the real change in the conformational entropy as it does 765 not account for the entropy change associated with motional modes on time scales longer than c (it only takes into 766 account the O2 values). The estimation of the change in the backbone conformational entropy was carried out using 767 ∆𝑆௖௢௡௙ൌ𝑅 ∑ ln ൤ ଵିை಺వభ ೒೗೤೎ೌ೟೐೏,೔ మ ଵିை಺వభ,೔ మ ൨௜ (5) 768 where O2 is the order parameter for each residue in native (I91)1 (𝑂ூଽଵ,௜ ଶ ) and in glycated (I91)1 (𝑂ூଽଵ ௚௟௬௖௔௧௘ௗ,௜ ଶ ), and 769 the sum runs over all residues92. 770 771 NMR study of the formation of MOLD on I91 772 To assess the formation of the MG-derived crosslinking MOLD on MG-treated I91, we first assigned the 1H and 773 the 13C-NMR chemical shifts of the imidazole ring of sMOLD ( Table S1). This was achieved using a solution 774 containing commercial MOLD (Iris Biotech) and collecting the co rresponding 1D- 1H, 13C-HSQC, 13C-edited-775 HSQC, 1H,1H-TOCSY, and 13C-HMBC spectra. Lastly, the obtained NMR data was used to identify the formation 776 of MOLD on MG-treated I91, which was achieved collecting the 13C-HMBC spectrum of a solution containing non-777 labeled I91 that was previously incubated with 13C-MG (Sigma-Aldrich). 778 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 25 779 Western blotting 780 Purified recombinant proteins were separated by 12% SDS-PAGE. P roteins were transferred onto polyvinylidene 781 difluoride (PVDF) membranes (Biorad) and blocked in 5% non-fat dried milk solution in Tris-Buffered Saline with 782 0.1% Tween-20. Presence of glycation products was tested by 2 h incubation at RT with anti-CEL primary antibody 783 (1:400, Cosmo Bio), followed by 1 h incubation at RT with HRP-c onjugated anti-mouse secondary antibody 784 (1:5000, ThermoFisher Scientific). Bands were visualized using ECL Western Blot Reagents (Cytiva). 785 786 Analysis of recombinant (I91)1 by LC-MS/MS 787 Around 50 μg of protein were subjected to in-filter reduction a nd alkylation using iodoacetamide followed by 788 chymotrypsin digestion (Nanose p Centrifugal Devices with Omega Membrane-30K, PALL), and the resulting 789 peptides were desalted by C18 OMIX tips according to the manufacturer’s protocol, after which all fractions were 790 dried under vacuum before MS analysis. MS analyses were perform ed as for native titin samples. To search for 791 crosslinked lysines in the recombinant I91 titin sequence, we built a list of all possible MOLD-crosslinked 792 combinations by proximity of lysine residues (K6-K55, K35-K79 and K37-K85) excluding combinations that would 793 render the same mass as missed cleavages of non-modified peptid es (Table S10). Different charge states for each 794 combination were manually analyzed applying layouts with the QualBrowser 4.5.474.0 program (Thermo Xcalibur) 795 by monitoring the retention time colocalization of 8 diagnostic fragments (according to the corresponding 796 sequence), in the m/z window of the parental ion along the enti re gradient. Scans containing the coeluting 797 diagnostic-fragments for each combination were manually inspect ed to assign the remaining expected fragments 798 resulting from the breaking of the crosslinked constructs, incl uding the presence of signals belonging to expected 799 neutral losses (ammonia loss, water loss and carbonyl-group loss generating b-derived “a” fragmentation series). 800 801 Small-angle X-ray scattering 802 Solutions containing 140 μM (I91)8 that had been incubated in the presence or absence of MG were dialyzed against 803 mili-Q water. Afterwards, these solutions were used for SAXS st udies, which were carried out on a Xeuss 2.0 804 instrument (Xenocs) equipped with a microfocus Cu Kα source (λ 1.54 Å) and a Pilatus 300 k detector (Dectris). 805 The distance between the detector and the sample was calibrated using silver behenate, and samples were measured 806 using sample-detector distances of 370 mm. The measurements were carried out for 2 h under vacuum at 25ºC using 807 a BioCUBE device (Xenocs), equipped with a temperature-controll ed flow-through quartz capillary (1.9 mm 808 width). In addition, we also collected the scattering curve for milli-Q water, which was subtracted from the signal 809 of the protein solutions. RAW software 93 was used for data processing and analysis. GNOM was used to co mpute 810 the pair-distance distribution functions (Pr)94. 811 812 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 26 Purification of titin from mouse cardiac tissue 813 Mouse cardiac tissue obtained by cryopulverization of frozen he arts in liquid nitrogen was used to extract titin 814 following the protocol from53 with minor modifications. Buffer volumes were adjusted accordi ng to tissue weight 815 (1mL per g of tissue). ~ 200 mg of tissue were washed 4 times w ith 3 volumes of homogenization buffer (1 mM 816 NaHCO3 pH 7, 50 mM KCl, 5 mM EGTA, 0.01 % NaN3) supplemented with protease inhibitors (1 mM PMSF, 0.1 817 mM leupeptin, 0.02 mM E-64, all from Sigma-Aldrich) and 20 µg/mL trypsin inhibitor (Roche). Between washes, 818 fibers were pelleted by centrifugation (2,000 g, 10 min, 4ºC). Then, fibers were resuspended in 2 volumes of 819 extraction buffer (10 mM imidazole pH 7, 900 mM KCl, 2mM EGTA, 2 mM MgCl2, 0.01 % NaN3) supplemented 820 with protease inhibitors (1.5 mM PMSF, 0.2 mM leupeptin, 0.04 m M E-64, 40 µg/mL trypsin inhibitor) and 821 homogenized with a plastic microhomogenizer (Takara BioMasher) on ice. After homogenization, the tissue debris 822 was pelleted by centrifugation (30 min, 20,000 g, 4ºC) and the supernatant was diluted 4 times in precipitation 823 buffer (0.1 mM NaHCO 3 pH 7, 0.1 mM EGTA) supplemented with protease inhibitors (0.05 mM leupeptin, 0.01 824 mM E-64) and incubated for 1 h at 4ºC in a spinning wheel. Afte r incubation, samples were centrifuged (30 min, 825 20,000 g, 4ºC) and the supernatant was diluted 5 times with precipitation buffer and incubated for 40 min at 4ºC in 826 a spinning wheel. Finally, the supernatants were centrifuged (4 0 min, 10,000 g, 4ºC) to precipitate titin molecules 827 and pellets were carefully resuspended in storage buffer (30 mM potassium phosphate pH 7.0, 200 mM KCl) at a 828 ratio of 500 µl of buffer per 200 mg of cardiac tissue used as starting material. Resuspended titin molecules were 829 stored at 4ºC until use. 830 831 Single-molecule force-spectroscopy on native titin 832 The experiments were carried out on HaloTag-ligand-coated surfaces, prepared according to published protocols53. 833 Briefly, glass coverslips (15 mm, Ted Pella) were cleaned by se quential sonication (1% Hellmanex, acetone, 96% 834 ethanol; 30 min each), air-dried, activated by exposure to O₂ plasma for 15 min, and silanized by immersion for 20 835 min in 1% (v/v) 3-aminopropyltrimethoxysilane (APTMS, Sigma-Ald rich) in ethanol. Unreacted silane was 836 removed by ethanol rinses, then surfaces were air-dried, cured at 100°C for ≥1 h, and stored in a desiccator until 837 use. For functionalization with HaloTag ligand, silanized cover slips were incubated overnight at RT in the dark 838 with 1 mM HaloTag succinimidyl ester (O4, Promega) in 50 mM borax buffer (pH 8.5). After incubation, coverslips 839 were washed with Milli-Q water, ai r-dried, and stored at 4°C in a humid chamber for up to one week. For titin 840 unfolding experiments, 20 µL Halo–titin solution was applied to a HaloTag-ligand-coated surface, incubated for 10 841 min, and washed with phosphate buffer. Single-molecule AFM force spectroscopy was performed in force–842 extension mode in 0.2 M phosphate buffer pH 7.4 containing 50 m M MG or control buffer without MG, using 843 MLCT Bio DC cantilevers (Bruker; nominal spring constant 0.01 N m⁻¹). Cantilevers were calibrated by the thermal 844 noise method71. Single titin proteins were picked up by pressing the tip against the coverslip for 1 second at 2 nN 845 contact force and unfolding was triggered by applying a ramp of 1.2 µm at a speed of 600 nm s⁻¹. Force–extension 846 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 27 curves were baseline-corrected and peaks corresponding to titin domain unfolding were identified. The associated 847 contour length increments (ΔLc) were estimated using the worm-like chain model50. 848 849 Monte Carlo simulations 850 To estimate the effect of glycation on titin stiffness we applied Monte Carlo simulations to a model of titin’s I-band 851 containing equivalent Ig domains (104 and 48 for the N2BA and N 2B titin isoforms) and two entropic regions 852 (PEVK and N2Bus) 43. Associated contour lengths and Kuhn lengths are included in Table S3. To calculate the 853 reduction in contour length by crosslinking modifications of th e PEVK, we aligned the 31 domains of the PEVK 854 region of human N2BA titin (Uniprot: Q8WZ42) using Clustal Omega and considered the possibility that the first 855 and last lysine in each domain can form a crosslink. To calcula te the associated reduction in contour length, we 856 subtracted the positions of both lysines from the number of amino acid residues present in the domain, considering 857 a contour length per amino acid of 0.4 nm 51. Next, we added the contour length of MOLD, which we estimated to 858 be 1.6 nm. As an example, for a PEVK domain that is 27 amino acid long and contains lysines at positions 6 and 859 27, the contour length upon crosslinking glycation would be (27-21) × 0.4 + 1.6 = 4 nm. Since the original contour 860 length is 27 × 0.4 = 10.8 nm, the contour length is reduced by 63% in this example. For the simulations, we 861 implemented an oscillating force protocol with triangular waves at a frequency of 1 Hz and 10 ms of simulation 862 step for 1 h and measured the resulting length of titin. For each condition, we ran 10 independent simulations. Next, 863 we averaged titin length at times longer than 300 s to ensure t hat steady state had been reached. Ratios between 864 steady state lengths at different conditions were calculated an d plotted as heatmaps. We used the Freely-Jointed 865 Chain model95 to estimate lengths from contour length values. Unfolding even ts were considered as all-or-none 866 events and their probability was calculated according to Bell’s model74. Parameters governing (un)folding kinetics 867 of Ig domains are included in Tables S4 and S5. 868 869 Statistical analysis 870 The statistical analyses were conducted with either IGOR Pro or GraphPad Prism 8. The data are presented with 871 the mean value either accompanied by the standard error of the mean (SEM) or standard deviation (SD), as specified. 872 The statistical significance was determined by using Student T- tests. If experimental data did not follow a normal 873 distribution, Mann-Whitney test was applied instead. Significan ce was defined as p-value <0.05 (*), <0.01(**), 874 <0.001 (***). 875 876

877 1 Mammoto, T., Mammoto, A. & Ingber, D. E. Mechanobiology and d evelopmental control. Annu Rev Cell 878 Dev Biol 29, 27-61, (2013). . 879 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 28 2 Paszek, M. J., Zahir, N., John son, K. R., Lakins, J. N., Rozenberg, G. I., Gefen, A., . . . Weaver, V. M. 880 Tensional homeostasis and the malignant phenotype. Cancer Cell 8, 241-254, (2005). 881 . 882 3 Borlaug Barry, A., Sharma, K., Shah Sanjiv, J. & Ho Jennifer, E. Heart Failure With Preserved Ejection 883 Fraction. JACC 81, 1810-1834, (2023). . 884 4 Zuela-Sopilniak, N. & Lammerdi ng, J. Can’t handle the stress? Mechanobiology and disease. Trends in 885 Molecular Medicine 28, 710-725, (2022). . 886 5 Lehman, S. J., Crocini, C. & L einwand, L. A. Targeting the sarcomere in inherited cardiomyopathies. 887 Nature Reviews Cardiology 19, 353-363, (2022). . 888 6 Falcao-Pires, I., Hamdani, N., Borbely, A., Gavina, C., Schalkwijk, C. G., van der Velden, J., . . . Paulus, W. J. 889 Diabetes mellitus worsens diastolic left ventricular dysfunction in aortic stenosis through altered 890 myocardial structure and cardiomyocyte stiffness. Circulation 124, 1151-1159, (2011). 891 . 892 7 van Heerebeek, L., Hamdani, N. , Handoko, M. L., Falcao-Pires, I., Musters, R. J., Kupreishvili, K., . . . 893 Paulus, W. J. Diastolic stiffness of the failing diabetic heart: importance of fibrosis, advanced glycation 894 end products, and myocyte resting tension. Circulation 117, 43-51, (2008). 895 . 896 8 Prenner, S. B. & Chirinos, J. A. Arterial stiffness in diabet es mellitus. Atherosclerosis 238, 370-379, 897 (2015). . 898 9 Loescher, C. M., Freundt, J. K. , Unger, A., Hessel, A. L., Kuhn, M., Koser, F. & Linke, W. A. Titin governs 899 myocardial passive stiffness with major support from microtubules and actin and the extracellular 900 matrix. Nat Cardiovasc Res 2, 991-1002, (2023). . 901 10 Bielajew, B. J., Hu, J. C. & Athanasiou, K. A. Collagen: quantification, biomechanics, and role of minor 902 subtypes in cartilage. Nat Rev Mater 5, 730-747, (2020). 903 . 904 11 Linke, W. A. & Hamdani, N. Gi gantic business: titin properties and function through thick and thin. Circ 905 Res 114, 1052-1068, (2014). . 906 12 Granzier, H. L. & Labeit, S. Discovery of Titin and Its Role in Heart Function and Disease. Circ Res 136, 907 135-157, (2025). . 908 13 Martinez-Martin, I., Silva-Ro jas, R. & Alegre-Cebollada, J. Mending the Achilles heels of titin in cardiac 909 and musculoskeletal disease. Biophysical Reviews In press, (2026). 910 14 Alegre-Cebollada, J. Protein nanomechanics in biological context. Biophys Rev 13, 435-454, (2021). 911 . 912 15 Loescher, C. M., Breitkreuz, M., Li, Y., Nickel, A., Unger, A., Dietl, A., . . . Linke, W. A. Regulation of titin-913 based cardiac stiffness by unfolded domain oxidation (UnDOx). Proc Natl Acad Sci U S A 117, 24545-914 24556, (2020). . 915 16 Binek, A., Castans, C., Jorge , I., Bagwan, N., Rodriguez, J. M., Fernandez-Jimenez, R., . . . Vazquez, J. 916 Oxidative Post-translational Protein Modifications upon Ischemia/Reperfusion Injury. Antioxidants 917 (Basel) 13, (2024). . 918 17 Uceda, A. B., Marino, L., Casasnovas, R. & Adrover, M. An ov erview on glycation: molecular mechanisms, 919 impact on proteins, pathogenesis, and inhibition. Biophys Rev 16, 189-218, (2024). 920 . 921 18 Fournet, M., Bonte, F. & De smouliere, A. Glycation Damage: A Possible Hub for Major 922 Pathophysiological Disorders and Aging. Aging Dis 9, 880-900, (2018). 923 . 924 19 Martin, M. S., Jacob-Dolan, J. W., Pham, V. T. T., Sjoblom, N. M. & Scheck, R. A. The chemical language of 925 protein glycation. Nat Chem Biol 21, 324-336, (2025). 926 . 927 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 29 20 Schalkwijk, C. G. & Stehouwer , C. D. A. Methylglyoxal, a Highly Reactive Dicarbonyl Compound, in 928 Diabetes, Its Vascular Complications, and Other Age-Related Diseases. Physiol Rev 100, 407-461, (2020). 929 . 930 21 Allaman, I., Belanger, M. & Magistretti, P. J. Methylglyoxal , the dark side of glycolysis. Front Neurosci 9, 931 23, (2015). . 932 22 Thornalley, P. J. Dicarbonyl intermediates in the maillard reaction. Ann N Y Acad Sci 1043, 111-117, 933 (2005). . 934 23 Lai, S. W. T., Lopez Gonzalez, E. J., Zoukari, T., Ki, P. & Shuck, S. C. Methylglyoxal and Its Adducts: 935 Induction, Repair, and Association with Disease. Chem Res Toxicol 35, 1720-1746, (2022). 936 . 937 24 Ahmed, M. U., Brinkmann Frye, E., Degenhardt, T. P., Thorpe, S. R. & Baynes, J. W. N-epsilon-938 (carboxyethyl)lysine, a product of the chemical modification of proteins by methylglyoxal, increases with 939 age in human lens proteins. Biochem J 324 ( Pt 2), 565-570, (1997). 940 . 941 25 Brinkmann, E., Wells-Knecht, K. J., Thorpe, S. R. & Baynes, J. W. Characterization of an imidazolium 942 compound formed by reaction of methylglyoxal and Nα-hippuryllysine. Journal of the Chemical Society, 943 Perkin Transactions 1, 2817-2818, (1995). . 944 26 Verzijl, N., DeGroot, J., Ben , Z. C., Brau-Benjamin, O., Maroudas, A., Bank, R. A., . . . TeKoppele, J. M. 945 Crosslinking by advanced glycation end products increases the stiffness of the collagen network in 946 human articular cartilage: a possible mechanism through which age is a risk factor for osteoarthritis. 947 Arthritis Rheum 46, 114-123, (2002). . 948 27 Norton, G. R., Candy, G. & Woodiwiss, A. J. Aminoguanidine P revents the Decreased Myocardial 949 Compliance Produced by Streptozotocin-Induced Diabetes Mellitus in Rats. Circulation 93, 1905-1912, 950 (1996). . 951 28 Sims, T. J., Rasmussen, L. M. , Oxlund, H. & Bailey, A. J. The role of glycation cross-links in diabetic 952 vascular stiffening. Diabetologia 39, 946-951, (1996). 953 . 954 29 Alegre-Cebollada, J., Kosuri, P., Giganti, D., Eckels, E., R ivas-Pardo, J. A., Hamdani, N., . . . Fernandez, J. 955 M. S-glutathionylation of cryptic cysteines enhances titin elasticity by blocking protein folding. Cell 156, 956 1235-1246, (2014). . 957 30 Targosz-Korecka, M., Jaglarz, M., Malek-Zietek, K. E., Gregorius, A., Zakrzewska, A., Sitek, B., . . . 958 Szymonski, M. AFM-based detection of glycocalyx degradation and endothelial stiffening in the db/db 959 mouse model of diabetes. Sci Rep 7, 15951, (2017). 960 . 961 31 Le Master, E., Fancher, I. S., Lee, J. & Levitan, I. Compara tive analysis of endothelial cell and sub-962 endothelial cell elastic moduli in young and aged mice: Role of CD36. Journal of Biomechanics 76, 263-963 268, (2018). . 964 32 Lieber, S. C., Aubry, N., Pai n, J., Diaz, G., Kim, S. J. & Vatner, S. F. Aging increases stiffness of cardiac 965 myocytes measured by atomic force microscopy nanoindentation. Am J Physiol Heart Circ Physiol 287, 966 H645-651, (2004). . 967 33 Hopf, A. E., Andresen, C., Ko tter, S., Isic, M., Ulrich, K., Sahin, S., . . . Kruger, M. Diabetes-Induced 968 Cardiomyocyte Passive Stiffening Is Caused by Impaired Insulin-Dependent Titin Modification and Can 969 Be Modulated by Neuregulin-1. Circ Res 123, 342-355, (2018). 970 . 971 34 Papadaki, M., Holewinski, R. J ., Previs, S. B., Martin, T. G., Stachowski, M. J., Li, A., . . . Kirk, J. A. Diabetes 972 with heart failure increases methylglyoxal modifications in the sarcomere, which inhibit function. JCI 973 Insight 3, (2018). . 974 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 30 35 Ruiz-Meana, M., Minguet, M., Bou-Teen, D., Miro-Casas, E., C astans, C., Castellano, J., . . . Garcia-975 Dorado, D. Ryanodine Receptor Glycation Favors Mitochondrial Damage in the Senescent Heart. 976 Circulation 139, 949-964, (2019). . 977 36 Coleman, D. L. Diabetes-obesity syndromes in mice. Diabetes 31, 1-6, (1982). 978 . 979 37 Huxley, H. E. The mechanism of muscular contraction. Science 164, 1356-1365, (1969). 980 . 981 38 Han, S.-W., Kolb, J., Farman, G. P., Gohlke, J. & Granzier, H. L. Glycerol storage increases passive stiffness 982 of muscle fibers through effects on titin extensibility. Journal of General Physiology 157, (2025). 983 . 984 39 Wagner, F. A., Loescher, C. M ., Unger, A., Kühn, M., Klotz, A. J., Liashkovich, I., . . . Linke, W. A. Direction-985 dependent contributions of cardiac myofilament networks to myocardial passive stiffness reveal a major 986 disparity for titin. Basic Research in Cardiology, (2025). . 987 40 Li, H., Linke, W. A., Oberhau ser, A. F., Carrion-Vazquez, M., Kerkvliet, J. G., Lu, H., . . . Fernandez, J. M. 988 Reverse engineering of the giant muscle protein titin. Nature 418, 998-1002, (2002). 989 . 990 41 Rivas-Pardo, J. A., Eckels, E . C., Popa, I., Kosuri, P., Linke, W. A. & Fernandez, J. M. Work Done by Titin 991 Protein Folding Assists Muscle Contraction. Cell Rep 14, 1339-1347, (2016). 992 . 993 42 Giganti, D., Yan, K., Badilla, C. L., Fernandez, J. M. & Alegre-Cebollada, J. Disulfide isomerization 994 reactions in titin immunoglobulin domains enable a mode of protein elasticity. Nat Commun 9, 185, 995 (2018). . 996 43 Herrero-Galan, E., Martinez-Martin, I., Sanchez-Gonzalez, C. , Vicente, N., Bonzon-Kulichenko, E., Calvo, 997 E., . . . Alegre-Cebollada, J. Basal oxidation of conserved cysteines modulates cardiac titin stiffness and 998 dynamics. Redox Biol 52, 102306, (2022). . 999 44 Popa, I., Kosuri, P., Alegre- Cebollada, J., Garcia-Manyes, S. & Fernandez, J. M. Force dependency of 1000 biochemical reactions measured by single-molecule force-clamp spectroscopy. Nat Protoc 8, 1261-1276, 1001 (2013). . 1002 45 Carl, P., Kwok, C. H., Manderson, G., Speicher, D. W. & Disc her, D. E. Forced unfolding modulated by 1003 disulfide bonds in the Ig domains of a cell adhesion molecule. Proc Natl Acad Sci U S A 98, 1565-1570, 1004 (2001). . 1005 46 Alegre-Cebollada, J., Badilla, C. L. & Fernandez, J. M. Isopeptide bonds block the mechanical extension of 1006 pili in pathogenic Streptococcus pyogenes. J Biol Chem 285, 11235-11242, (2010). 1007 . 1008 47 Lipari, G. & Szabo, A. Model- free approach to the interpretation of nuclear magnetic resonance 1009 relaxation in macromolecules. 1. Theory and range of validity. Journal of the American Chemical Society 1010 104, 4546-4559, (1982). . 1011 48 Li, Z., Raychaudhuri, S. & Wa nd, A. J. Insights into the local residual entropy of proteins provided by NMR 1012 relaxation. Protein Sci 5, 2647-2650, (1996). . 1013 49 Marino, L., Ramis, R., Casasn ovas, R., Ortega-Castro, J., Vilanova, B., Frau, J. & Adrover, M. Unravelling 1014 the effect of N(epsilon)-(carboxyethyl)lysine on the conformation, dynamics and aggregation propensity 1015 of alpha-synuclein. Chem Sci 11, 3332-3344, (2020). 1016 . 1017 50 Bustamante, C., Marko, J. F., Siggia, E. D. & Smith, S. Entr opic elasticity of lambda-phage DNA. Science 1018 265, 1599-1600, (1994). . 1019 51 Ainavarapu, S. R., Brujic, J., Huang, H. H., Wiita, A. P., L u, H., Li, L., . . . Fernandez, J. M. Contour length 1020 and refolding rate of a small protein controlled by engineered disulfide bonds. Biophys J 92, 225-233, 1021 (2007). . 1022 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 31 52 Rambo, R. P. & Tainer, J. A. Characterizing flexible and intrinsically unstruc tured biological 1023 macromolecules by SAS using the Porod-Debye law. Biopolymers 95, 559-571, (2011). 1024 . 1025 53 Rivas-Pardo, J. A., Li, Y., M ártonfalvi, Z., Tapia-Rojo, R., Unger, A., Fernández-Trasancos, Á., . . . Alegre-1026 Cebollada, J. A HaloTag-TEV genetic cassette for mechanical phenotyping of proteins from tissues. 1027 Nature Communications 11, 2060, (2020). . 1028 54 Rief, M., Gautel, M., Oesterh elt, F., Fernandez, J. M. & Gaub, H. E. Reversible unfolding of individual titin 1029 immunoglobulin domains by AFM. Science 276, 1109-1112, (1997). 1030 . 1031 55 Schlierf, M., Li, H. & Fernan dez, J. M. The unfolding kinetics of ubiquitin captured with single-molecule 1032 force-clamp techniques. Proc Natl Acad Sci U S A 101, 7299-7304, (2004). 1033 . 1034 56 Kroemer, G., Lopez-Otin, C., Madeo, F. & de Cabo, R. Carbotoxicity-Noxious Effects of Carbohydrates. 1035 Cell 175, 605-614, (2018). . 1036 57 Methawasin, M. & Granzier, H. Softening the Stressed Giant T itin in Diabetes Mellitus. Circ Res 123, 315-1037 317, (2018). . 1038 58 Mellor, K. M., Varma, U., Kout sifeli, P., Curl, C. L., Janssens, J. V., Daniels, L. J., . . . Delbridge, L. M. D. 1039 Targeted glycophagy ATG8 therapy reverses diabetic heart disease in mice and in human engineered 1040 cardiac tissues. Nat Cardiovasc Res 4, 1487-1500, (2025). 1041 . 1042 59 Trujillo, M. N. & Galligan, J. J. Reconsidering the role of protein glycation in disease. Nature Chemical 1043 Biology 19, 922-927, (2023). . 1044 60 Di Sanzo, S., Spengler, K., L eheis, A., Kirkpatrick, J. M., Randler, T. L., Baldensperger, T., . . . Heller, R. 1045 Mapping protein carboxymethylation sites provides insights into their role in proteostasis and cell 1046 proliferation. Nat Commun 12, 6743, (2021). . 1047 61 Elosegui-Artola, A., Trepat, X. & Roca-Cusachs, P. Control o f Mechanotransduction by Molecular Clutch 1048 Dynamics. Trends in Cell Biology 28, 356-367, (2018). 1049 . 1050 62 Leitner, A., Walzthoeni, T., Kahraman, A., Herzog, F., Rinner, O., Beck, M. & Aebersold, R. Probing native 1051 protein structures by chemical cross-linking, mass spectrometry, and bioinformatics. Mol Cell Proteomics 1052 9, 1634-1649, (2010). . 1053 63 Trnka, M. J., Baker , P. R., Robinson, P. J., Burlingame, A. L. & Chalkley, R. J. Matching cross-linked peptide 1054 spectra: only as good as the worse identification. Mol Cell Proteomics 13, 420-434, (2014). 1055 . 1056 64 Delligatti, C. E. & Kirk, J. A. in Vitamins and Hormones Vol. 125 (ed Gerald Litwack) 47-88 (Academic 1057 Press, 2024). 1058 65 LeWinter, M. M., Taatjes, D., Ashikaga, T., Palmer, B., Bishop, N., VanBuren, P., . . . Zile, M. Abundance, 1059 localization, and functional correlates of the advanced glycation end-product carboxymethyl lysine in 1060 human myocardium. Physiological Reports 5, e13462, (2017). 1061 . 1062 66 Janssens, J. V., Ma, B., Brimb le, M. A., Van Eyk, J. E., Delbridge, L. M. D. & Mellor, K. M. Cardiac troponins 1063 may be irreversibly modified by glycation: novel potential mechanisms of cardiac performance 1064 modulation. Scientific Reports 8, 16084, (2018). . 1065 67 Papadaki, M., Kampaengsri, T., Barrick, S. K., Campbell, S. G., von Lewinski, D., Rainer, P. P., . . . Kirk, J. A. 1066 Myofilament glycation in diabetes reduces contractility by inhibiting tropomyosin movement, is rescued 1067 by cMyBPC domains. J Mol Cell Cardiol 162, 1-9, (2022). 1068 . 1069 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 32 68 Lewis, C. T. A., Moreno-Justi cia, R., Savoure, L., Calvo, E., Bak, A., Laitila, J., . . . Ochala, J. Dysregulated 1070 skeletal muscle myosin super-relaxation and energetics in male participants with type 2 diabetes 1071 mellitus. Diabetologia 68, 1836-1850, (2025). . 1072 69 Shao, C. H., Capek, H. L., Patel, K. P., Wang, M., Tang, K., DeSouza, C., . . . Bidasee, K. R. Carbonylation 1073 Contributes to SERCA2a Activity Loss and Diastolic Dysfunction in a Rat Model of Type 1 Diabetes. 1074 Diabetes 60, 947-959, (2011). . 1075 70 Friedemann, T. E. The action of alkali and hydrogen peroxide on glyoxals. Journal of Biological Chemistry 1076 73, 331-334, (1927). . 1077 71 Hutter, J. L. & Bechhoefer, J . Calibration of atomic-force microscope tips. Review of Scientific 1078 Instruments 64, 1868-1873, (1993). . 1079 72 Schillers, H., Rianna, C., Schape, J., Luque, T., Doschke, H ., Walte, M., . . . Radmacher, M. Standardized 1080 Nanomechanical Atomic Force Microscopy Procedure (SNAP) for Measuring Soft and Biological Samples. 1081 Sci Rep 7, 5117, (2017). . 1082 73 Goncalves-Rodrigues, P., Alme ida-Coelho, J., Goncalves, A., Amorim, F., Leite-Moreira, A. F., Stienen, G. 1083 J. M. & Falcao-Pires, I. In Vitro Assessment of Cardiac Function Using Skinned Cardiomyocytes. J Vis Exp, 1084 (2020). . 1085 74 Bell, G. I. Models for the specific adhesion of cells to cel ls. Science 200, 618-627, (1978). 1086 . 1087 75 Kosuri, P., Alegre-Cebollada, J., Feng, J., Kaplan, A., Ingles-Prieto, A., Badilla, C. L., . . . Fernandez, J. M. 1088 Protein folding drives disulfide formation. Cell 151, 794-806, (2012). 1089 . 1090 76 Piotto, M., Saudek, V. & Skle nar, V. Gradient-tailored excitation for single-quantum NMR spectroscopy 1091 of aqueous solutions. J Biomol NMR 2, 661-665, (1992). 1092 . 1093 77 Wishart, D. S., Bigam, C. G., Yao, J., Abildgaard, F., Dyson, H. J., Oldfield, E., . . . Sykes, B. D. 1H, 13C and 1094 15N chemical shift referencing in biomolecular NMR. J Biomol NMR 6, 135-140, (1995). 1095 . 1096 78 Delaglio, F., Grzesiek, S., Vu ister, G. W., Zhu, G., Pfeifer, J. & Bax, A. NMRPipe: a multidimensional 1097 spectral processing system based on UNIX pipes. J Biomol NMR 6, 277-293, (1995). 1098 . 1099 79 Bartels, C., Xia, T. H., Bill eter, M., Guntert, P. & Wuthrich, K. The program XEASY for computer-supported 1100 NMR spectral analysis of biological macromolecules. J Biomol NMR 6, 1-10, (1995). 1101 . 1102 80 Goddard, T. & Kneller, D. Spa rky 3. University of California, San Francisco, USA. There is no 1103 corresponding record for this reference, (2008). 1104 81 Improta, S., Politou, A. S. & Pastore, A. Immunoglobulin-like modules from titin I-band: extensible 1105 components of muscle elasticity. Structure 4, 323-337, (1996). 1106 . 1107 82 Tamiola, K. & Mulder, F. A. Usi ng NMR chemical shifts to calculate the propensity for structural order 1108 and disorder in proteins. Biochem Soc Trans 40, 1014-1020, (2012). 1109 . 1110 83 Shen, Y., Delaglio, F., Corni lescu, G. & Bax, A. TALOS+: a hybrid method for predicting protein backbone 1111 torsion angles from NMR chemical shifts. J Biomol NMR 44, 213-223, (2009). 1112 . 1113 84 Hafsa, N. E., Arndt, D. & Wis hart, D. S. CSI 3.0: a web server for identifying secondary and super-1114 secondary structure in proteins using NMR chemical shifts. Nucleic Acids Res 43, W370-377, (2015). 1115 . 1116 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 33 85 Farrow, N. A., Muhandiram, R., Singer, A. U., Pascal, S. M., Kay, C. M., Gish, G., . . . Kay, L. E. Backbone 1117 dynamics of a free and phosphopeptide-complexed Src homology 2 domain studied by 15N NMR 1118 relaxation. Biochemistry 33, 5984-6003, (1994). . 1119 86 Dosset, P., Hus, J. C., Blackledge, M. & Marion, D. Efficien t analysis of macromolecular rotational 1120 diffusion from heteronuclear relaxation data. J Biomol NMR 16, 23-28, (2000). 1121 . 1122 87 Lee, L. K., Rance, M., Chazin , W. J. & Palmer, A. G., 3rd. Rotational diffusion anisotropy of proteins from 1123 simultaneous analysis of 15N and 13C alpha nuclear spin relaxation. J Biomol NMR 9, 287-298, (1997). 1124 . 1125 88 Schrodinger, LLC. The PyMOL Molecular Graphics System, Version 1.8 (2015). 1126 89 Kroenke, C. D., Rance, M. & P almer, A. G. Variability of the 15N Chemical Shift Anisotropy in Escherichia 1127 coli Ribonuclease H in Solution. Journal of the American Chemical Society 121, 10119-10125, (1999). 1128 . 1129 90 Jarymowycz, V. A. & Stone, M. J. Fast time scale dynamics of protein backbones: NMR relaxation 1130 methods, applications, and functional consequences. Chem Rev 106, 1624-1671, (2006). 1131 . 1132 91 Akke, M., Brueschweiler, R. & Palmer, A. G., III. NMR order parameters and free energy: an analytical 1133 approach and its application to cooperative calcium(2+) binding by calbindin D9k. Journal of the 1134 American Chemical Society 115, 9832-9833, (1993). . 1135 92 Li, D. W. & Bruschweiler, R. A dictionary for protein side-c hain entropies from NMR order parameters. J 1136 Am Chem Soc 131, 7226-7227, (2009). . 1137 93 Hopkins, J. B., Gillilan, R. E. & Skou, S. BioXTAS RAW: impr ovements to a free open-source program for 1138 small-angle X-ray scattering data reduction and analysis. J Appl Crystallogr 50, 1545-1553, (2017). 1139 . 1140 94 Konarev, P. V., Petoukhov, M. V., Volkov, V. V. & Svergun, D. I. ATSAS 2.1, a program package for small-1141 angle scattering data analysis. Journal of Applied Crystallography 39, 277-286, (2006). 1142 . 1143 95 Smith, S. B., Finzi, L. & Bustamante, C. Direct mechanical m easurements of the elasticity of single DNA 1144 molecules by using magnetic beads. Science 258, 1122-1126, (1992). 1145 . 1146 1147 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 34 1148 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 35 Figure 1. Glycation stiffens cardiomyocytes. (A) Hypothesis of this work: protein glycation contributes to 1149 cardiomyocyte stiffening in situation of glycative stress inclu ding diabetes. (B) Schematic representation of a 1150 contracted and relaxed I-band in half a sarcomere (not to scale). Titin is colored in yellow, while other sarcomeric 1151 proteins appear in gray. Titin Ig and fibronectin domains are r epresented as filled circles, and the approximate 1152 positions of the mostly unstructur ed N2Bus and PEVK segments ar e indicated. The length of the mechanically 1153 active I-band and the beginning of the A-band are delimited by arrows. Please note that in the relaxed sarcomere, 1154 unstructured regions are extended, and a fraction of Ig domains is unfolded. (C) Difference in spectral counts 1155 (#Spectral counts) of peptides containing glycated residues alon g the titin sequence (x-axis) resulting from the 1156 hearts of 5 control and 5 ob/ob mice (2 chromatographic runs per sample). (D) Difference in spectral counts of 1157 peptides containing glycated residues between diabetic myocardi um samples (n=2) and control myocardium 1158 samples (n=2). In C and D, the pink area indicates values above and below 3 x SD of control datasets where no 1159 differences in glycated peptides are expected (Figure S1). (E) Representation of AFM nanoindentation experiments 1160 to characterize the effects of incubation with MG on transverse stiffness of skinned neonatal cardiomyocytes. (F) 1161 Representative approach force-distance AFM curves recorded for untreated (-MG, black) and MG-treated (+MG, 1162 red) cardiomyocytes. (G) Quantification of Young’s moduli obtained by AFM nanoindentation. Each dot represents 1163 the median Young’s modulus obtained for an individual cell, results obtained with n = 13 for -MG condition and n 1164 = 11 for +MG condition, error bars represent SD. (H) Stretch protocol used to measure the passive force of 1165 cardiomyocytes in the longitudinal direction. A single skinned cardiomyocyte is mounted between a force sensor 1166 and a motor (inset). The cell is stretched from 1.0 to 1.3 times its initial length (L0) in 6 steps, and the passive force 1167 is recorded before and after 30-minute incubation with phosphoc reatine-free relaxing buffer including or not MG 1168 at RT. Force traces show average results from n = 8 cells per c ondition; individual force traces were normalized 1169 considering the peak force values at 1.3 L 0 before the incubation phase. (I) Changes in passive force measured at 1170 1.3 initial length (L 0) before and after incubation with and without MG. (J) Relative changes in passive stiffness 1171 after incubation with and without MG. Error bars represent SEM (n = 8). p-value <0.01(**), <0.001 (***). 1172 1173 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 36 1174 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 37 Figure 2. MG induces a high proportion of crosslinking modifica tions in the I91 domain of titin. (A) 1175 Representation of the reaction between I91 and MG leading to the formation of non-crosslinking (e.g. CEL; depicted 1176 in blue) and crosslinking (e.g. MOLD; depicted in red) AGEs in lysine residues. (B) Experimental groups to study 1177 the effects of glycation of the (I91) 8 polyprotein. Non-modified lysine residues are represented in o ne-letter code; 1178 modifications indicated as in (A). (C) Size-exclusion chromatograms of (I91) 8 protein preparations. Arrow marks 1179 elution volume of (I91)8. (D) MALDI-TOF/TOF spectra of pristine or glycated (I91)8 after incubation with 50 mM 1180 MG for 72 h at 37°C. (E-G) Representative single-molecule force-ramp traces to probe the m echanical unfolding 1181 of individual titin I91 domains in the three types of (I91) 8 preparations. Blue stars mark 23-27 nm events coming 1182 from unfolding of full length I91 domains. Red stars indicate s horter events (9-14 nm). (H-J) Bidimensional 1183 histograms showing frequency of events according to their size and force at which they occur for the three 1184 experimental conditions, using lax fingerprinting to select sin gle-molecule events. The number of events 1185 contributing to the data set are indicated. Monodimensional distributions of step sizes are shown on top of the 1186 bidimensional histograms. (K-M) Equivalent histograms to panels H-J obtained using strict fingerprinting to select 1187 single-molecule events. (N) Kinetics of appearance of 9-14 nm steps. The solid line is a f it to first order reaction 1188 kinetics. For the 0 h incubation time, the data for pristine pr otein was considered for the fit. The open symbol 1189 corresponds to the results from the sample to which MG was adde d and then immediately removed using FPLC. 1190 Error bars are SD of bootstrapping distributions75. PDF: probability density function. 1191 1192 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 38 1193 Figure 3. Characterization of glycation-induced modifications i n titin’s I91 domain by NMR. (A) Ribbon 1194 representation of the solution structure of I91 (PDB: 1TIT). The side chains of lysine residues are shown in purple. 1195 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 39 The double arrow indicates the distance between the ends of the globular structure. (B) Overlapping of the 1196 projections of the different H-H planes of the 1H,1H-TOCSY-HSQC spectra obtained for native (black) and glycated 1197 (I91)1 (red). The chemical shifts of the H -H cross-peaks corresponding to lysine residues are marked with a blue 1198 line that crosses the correlation peaks appearing due to coupling between the H with the other protons of the lysine 1199 side chains. The insert in the spectra shows a model representi ng the side chain of lysine. (C) Overlapping of the 1200 1H,13C-HMBC spectrum obtained for sMOLD (blue) and that obtained for (I91)1 treated with 13C-MG (red). As it 1201 happens in 1H,13C-HMBC spectra, all the peaks corresponding to the C-H one bond correlation are split into two 1202 different signals separated by their 1JC-H coupling constants. The chemical structure of MOLD is shown. (D) Overlay 1203 of the ¹⁵N-HSQC spectra for non-glycated (I91) 1 (black) and glycated (I91) 1 (red). (E) Effect of glycation on the 1204 intensity of the HSQC peaks, calculated as (I glycated/Inative)-1, where Inative is the resonance intensity of each peak of 1205 I91 (I’native) internally corrected by the intensity of the N-H cross peak corresponding to the side chain of N77 (see 1206 panel D) (Inative=I’native/IN77(N-H)), and Iglycated is also the corrected resonance intensity of each peak in glycated (I91)1. 1207 The bars corresponding to lysine residues are colored in green. (F) Backbone O2 values obtained from NMR 1208 relaxation data for non-glycated (I91) 1 (black) and glycated (I91) 1 (red). Gray shaded areas indicate the positions 1209 of lysine residues. Position of -strands following previously assigned annotations81 is shown as blue arrows at the 1210 top of the panel. (G) Differences between order parameters for residues in non-glyca ted (I91)1 and glycated (I91)1 1211 (ΔO2= O2I91 glycated - O2I91). The bars corresponding to lysine residues are colored in gre en. ( H ) C a r t o o n 1212 representations of the 3D structure of I91, which have been color-coded according to the ΔO2 values plotted in panel 1213 G (blue for ΔO2 > 0.3, cyan for 0.2 < ΔO2 ≤ 0.3, red for 0.1 < ΔO2 ≤ 0.2, salmon for 0.05 < ΔO2 ≤ 0.1, and yellow 1214 for 0.02 < ΔO2 ≤ 0.05. The image presents two views of the same structure, rotated 180º. Labels for the β-sheets are 1215 shown in blue. Images were generated with Pymol88. 1216 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 40 1217 Figure 4. Lysines in I91 are mechanically relevant glycation targets. (A) Schematic representation of a lysine 1218 residue reacting with pyruvic acid in the presence of NaBH3CN to form CEL. (B) Experimental workflow to block 1219 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 41 lysines by formation of CEL (routes 3 and 4), which prevents su bsequent MG glycation in route 4. Control 1220 incubations of the (I91) 8 are also indicated (routes 1 and 2). (C) SDS-PAGE analysis of (I91) 8 domains that have 1221 undergone the reactions shown in (B). Panel shows Coomassie blue staining (left) and WB analysis using anti-CEL 1222 antibodies (right). (D) Bidimensional histograms showing the frequency of the size of the unfolding steps and the 1223 force at which they occur in AFS traces selected with stringent fingerprinting for the four experimental conditions. 1224 The number of events contributing to the data set are indicated . Monodimensional distributions of step sizes are 1225 shown on top of the bidimensional histograms. (E-F) Quantification of the percentage of unfolding steps between 1226 9 and 14 nm in the two experimental arms. (G) Distribution of force versus step size for events obtained using MG-1227 incubated (I91)8 in AFS experiments. Solid lines are worm-like chain fits to the data. The contour length reduction 1228 in short steps is estimated to correspond to a covalent barrier trapping around 40 amino acids. (H) Structure of the 1229 I91 domain (PDB: 1TIT) highlighting lysine residues mutated out in (I91-KA)8. Image was generated with Pymol88. 1230 (I-J) (I91-KA)8 was incubated in the absence or presence of 50 mM MG for 72 h a t 37ºC and probed using single-1231 molecule AFS. Single-molecule traces were selected with the non-stringent fingerprinting criterion, and data are 1232 represented as in panels D and E. WT data from Figure 2. Error bars are SD of bootstrapping distributions75. PDF: 1233 probability density function. 1234 1235 1236 1237 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 42 1238 Figure 5: Identification of a MOLD crosslink in glycated I91 by LC-MS/MS. (A) Schematic representation of 1239 the experimental workflow. (B) Zoomed-in integrated survey scan between minutes 36 and 37, showing the doubly-1240 charged precursor ion at m/z 698.38 corresponding to a chymotry ptic dipeptide from I91 crosslinked by a MOLD 1241 connector. (C) Sequence of the chymotryptic dipeptide; the connector and lys ines 35 and 79 are highlighted in 1242 green. (D) Assigned fragmentation MS/MS spectrum corresponding to the pe ptide in (C), taking the long peptide 1243 sequence (QAANTKSAANL) as a base on which the short sequence pl us the crosslinking agent (MOLD-KL) are 1244 added. The figure shows the main fragmentation series (b and y) for the fragments singly- (+) or doubly-charged 1245 (++), containing (asterisks) or not the MOLD-KL delta-mass, demonstrating the exact location of crosslinking. The 1246 y´1 and b´1 ions would correspond to the breaking of the peptid e bond between the residues of the short sequence 1247 (KL), giving rise to the common fragment derived from leucine (y1 / y´1) and the fragment corresponding to the 1248 rest of the crosslinked structure (b´1). Black arrows indicate neutral losses of water or ammonium. Blue arrows 1249 mark the a-series ions derived from the neutral loss of the car bonyl group from the corresponding b-series ions, 1250 demonstrating the correct assignment of these ions. 1251 1252 1253 1254 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 43 1255 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint 44 Figure 6. Mechanics of glycated titin. (A) Native titin molecules containing a HaloTag insertion at the end of the 1256 I-band region are immobilized on a HaloTag-ligand derivatized glass surface and stretched using AFM in constant-1257 velocity mode. Glycation is induced in situ by adding 50 mM MG. Representative curves are shown. (B) Cumulative 1258 distribution of changes in contour length (ΔLc). (C) Distribution of unfolding forces of I91 domains after 72 h 1259 incubation of (I91)8 at 37ºC in the absence (top; n = 440 events from 3 independent experiments) and in the presence 1260 of 50 mM MG ( bottom; n = 436 9-14 nm events from 2 independent experiments). (D) Distribution of unfolding 1261 forces for (I91)8 incubated for 48 h at 50 ºC without (top; n = 323 from 4 independent experiments) or with (bottom; 1262 n = 418 from 4 independent experiments) pyruvic acid and NaBH 3CN. Solid lines in panels E and F are fits to the 1263 Bell-Evans model of force-activated reactions 55,74. (E) Representative unfolding-quench-probe AFS trace to study 1264 mechanical folding of glycated I91. A single glycated (I91) 8 polyprotein is subjected to 40 pN s -1 unfolding pulse, 1265 then force is quenched to 0 pN and finally increased again in a probe pulse. The folding fraction is calculated by 1266 comparing the number of unfolded domains in both ramps. In this example, 5 out of 5 9-14 nm steps and 2 out of 3 1267 23-27 nm steps refolded during the force quench. 40 pN force pulses are included for optimal sensitivity (see main 1268 text). (F,G) Folding fractions corresponding to the same protein preparation s a s i n ( C , D ) . L i n e s r e p r e s e n t 1269 exponential fits to the data. In all cases, n>60 for each time point. (H) Schematic representation of the PEVK region 1270 in N2BA titin, highlighting the position of lysine residues in the sequence of human N2BA titin (Uniprot: Q8WZ42). 1271 (I) Alignment of the 31 PEVK repeats of human N2BA titin (repeats a ccording to Uniprot: Q8WZ42). The most 1272 terminal lysine residues in each repeat are highlighted in red, while the remaining lysine residues are marked in 1273 pink. All other amino acids are shown in gray. (J) Distribution of contour length reduction in all PEVK repeats 1274 when their first and last lysine residues are crosslinked by MOLD. (K) Force protocol for Monte Carlo simulations. 1275 (L) Corresponding protein length of a virtual N2BA titin subjected to to the force protocol in (K). (M) Length of 1276 N2BA titin at 10 pN during the Monte Carlo simulations for non-modified (black), 100% glycated/0% crosslinked 1277 (blue) and 100% glycated/100% crosslinked (red) protein. Solid lines are the average of 10 independent simulations. 1278 Shaded areas represent SD. (N) Heatmap representing N2BA titin length at 10 pN peak force for Monte Carlo 1279 simulations at different glycation conditions, relative to the non-modified protein. Results are the average of 10 1280 independent simulations per condition. The simulations in panels M and N consider that the PEVK region can be 1281 glycated. Error bars are SD of bootstrapping distributions75. PDF: probability density function. 1282 1283 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 14, 2026. ; https://doi.org/10.64898/2026.02.13.705744doi: bioRxiv preprint